diff --git a/Makefile b/Makefile
index b67ecacd..36a451cc 100644
--- a/Makefile
+++ b/Makefile
@@ -1,6 +1,6 @@
-GPU=1
-CUDNN=1
-OPENCV=1
+GPU=0
+CUDNN=0
+OPENCV=0
 DEBUG=0
 
 ARCH= -gencode arch=compute_20,code=[sm_20,sm_21] \
@@ -10,7 +10,7 @@ ARCH= -gencode arch=compute_20,code=[sm_20,sm_21] \
       -gencode arch=compute_52,code=[sm_52,compute_52]
 
 # This is what I use, uncomment if you know your arch and want to specify
-ARCH=  -gencode arch=compute_52,code=compute_52
+# ARCH=  -gencode arch=compute_52,code=compute_52
 
 VPATH=./src/:./examples
 LIB=libdarknet.a
@@ -23,7 +23,7 @@ AR=ar
 ARFLAGS=-rv
 OPTS=-Ofast
 LDFLAGS= -lm -pthread 
-COMMON= -Iinclude/
+COMMON= -Iinclude/ -Isrc/
 CFLAGS=-Wall -Wfatal-errors 
 
 ifeq ($(DEBUG), 1) 
@@ -60,7 +60,7 @@ endif
 
 EXECOBJ = $(addprefix $(OBJDIR), $(EXECOBJA))
 OBJS = $(addprefix $(OBJDIR), $(OBJ))
-DEPS = $(wildcard include/darknet/*.h) Makefile
+DEPS = $(wildcard src/*.h) Makefile include/darknet.h
 
 all: obj backup results $(LIB) $(EXEC)
 
diff --git a/examples/art.c b/examples/art.c
index 53fe60d0..3f03f79f 100644
--- a/examples/art.c
+++ b/examples/art.c
@@ -1,9 +1,5 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/option_list.h"
-#include "darknet/blas.h"
-#include "darknet/classifier.h"
+#include "darknet.h"
+
 #include <sys/time.h>
 
 void demo_art(char *cfgfile, char *weightfile, int cam_index)
diff --git a/examples/captcha.c b/examples/captcha.c
index dce713ae..1de21fa3 100644
--- a/examples/captcha.c
+++ b/examples/captcha.c
@@ -1,6 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
+#include "darknet.h"
 
 void fix_data_captcha(data d, int mask)
 {
diff --git a/examples/cifar.c b/examples/cifar.c
index fbbea285..a05d0997 100644
--- a/examples/cifar.c
+++ b/examples/cifar.c
@@ -1,8 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/option_list.h"
-#include "darknet/blas.h"
+#include "darknet.h"
 
 void train_cifar(char *cfgfile, char *weightfile)
 {
diff --git a/examples/classifier.c b/examples/classifier.c
index 440ab21a..112b913d 100644
--- a/examples/classifier.c
+++ b/examples/classifier.c
@@ -1,10 +1,5 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/option_list.h"
-#include "darknet/blas.h"
-#include "darknet/classifier.h"
-#include "darknet/cuda.h"
+#include "darknet.h"
+
 #include <sys/time.h>
 #include <assert.h>
 
@@ -37,11 +32,7 @@ void train_classifier(char *datacfg, char *cfgfile, char *weightfile, int *gpus,
 #ifdef GPU
         cuda_set_device(gpus[i]);
 #endif
-        nets[i] = parse_network_cfg(cfgfile);
-        if(weightfile){
-            load_weights(&nets[i], weightfile);
-        }
-        if(clear) *nets[i].seen = 0;
+        nets[i] = load_network(cfgfile, weightfile, clear);
         nets[i].learning_rate *= ngpus;
     }
     srand(time(0));
diff --git a/examples/coco.c b/examples/coco.c
index 02e62455..bab507d9 100644
--- a/examples/coco.c
+++ b/examples/coco.c
@@ -1,12 +1,6 @@
-#include <stdio.h>
+#include "darknet.h"
 
-#include "darknet/network.h"
-#include "darknet/detection_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/box.h"
-#include "darknet/demo.h"
+#include <stdio.h>
 
 char *coco_classes[] = {"person","bicycle","car","motorcycle","airplane","bus","train","truck","boat","traffic light","fire hydrant","stop sign","parking meter","bench","bird","cat","dog","horse","sheep","cow","elephant","bear","zebra","giraffe","backpack","umbrella","handbag","tie","suitcase","frisbee","skis","snowboard","sports ball","kite","baseball bat","baseball glove","skateboard","surfboard","tennis racket","bottle","wine glass","cup","fork","knife","spoon","bowl","banana","apple","sandwich","orange","broccoli","carrot","hot dog","pizza","donut","cake","chair","couch","potted plant","bed","dining table","toilet","tv","laptop","mouse","remote","keyboard","cell phone","microwave","oven","toaster","sink","refrigerator","book","clock","vase","scissors","teddy bear","hair drier","toothbrush"};
 
diff --git a/examples/darknet.c b/examples/darknet.c
index 4592dc3e..31652b12 100644
--- a/examples/darknet.c
+++ b/examples/darknet.c
@@ -1,14 +1,9 @@
+#include "darknet.h"
+
 #include <time.h>
 #include <stdlib.h>
 #include <stdio.h>
 
-#include "darknet/parser.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
-#include "darknet/connected_layer.h"
-#include "darknet/convolutional_layer.h"
-
 extern void predict_classifier(char *datacfg, char *cfgfile, char *weightfile, char *filename, int top);
 extern void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh, float hier_thresh, char *outfile, int fullscreen);
 extern void run_voxel(int argc, char **argv);
diff --git a/examples/detector.c b/examples/detector.c
index 2c8234d8..0a31fc27 100644
--- a/examples/detector.c
+++ b/examples/detector.c
@@ -1,12 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/region_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/box.h"
-#include "darknet/demo.h"
-#include "darknet/option_list.h"
-#include "darknet/blas.h"
+#include "darknet.h"
 
 static int coco_ids[] = {1,2,3,4,5,6,7,8,9,10,11,13,14,15,16,17,18,19,20,21,22,23,24,25,27,28,31,32,33,34,35,36,37,38,39,40,41,42,43,44,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,67,70,72,73,74,75,76,77,78,79,80,81,82,84,85,86,87,88,89,90};
 
diff --git a/examples/dice.c b/examples/dice.c
index 7a2a1949..6a810492 100644
--- a/examples/dice.c
+++ b/examples/dice.c
@@ -1,6 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
+#include "darknet.h"
 
 char *dice_labels[] = {"face1","face2","face3","face4","face5","face6"};
 
diff --git a/examples/go.c b/examples/go.c
index 3e3d5f67..f4513574 100644
--- a/examples/go.c
+++ b/examples/go.c
@@ -1,9 +1,5 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/option_list.h"
-#include "darknet/blas.h"
-#include "darknet/data.h"
+#include "darknet.h"
+
 #include <unistd.h>
 
 int inverted = 1;
diff --git a/examples/lsd.c b/examples/lsd.c
index 61f8e8f2..85641ecd 100644
--- a/examples/lsd.c
+++ b/examples/lsd.c
@@ -1,8 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/blas.h"
+#include "darknet.h"
 
 /*
 void train_lsd3(char *fcfg, char *fweight, char *gcfg, char *gweight, char *acfg, char *aweight, int clear)
diff --git a/examples/nightmare.c b/examples/nightmare.c
index 3eda5fb6..9538238d 100644
--- a/examples/nightmare.c
+++ b/examples/nightmare.c
@@ -1,8 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/parser.h"
-#include "darknet/blas.h"
-#include "darknet/utils.h"
-#include "darknet/region_layer.h"
+#include "darknet.h"
 
 // ./darknet nightmare cfg/extractor.recon.cfg ~/trained/yolo-coco.conv frame6.png -reconstruct -iters 500 -i 3 -lambda .1 -rate .01 -smooth 2
 
diff --git a/examples/regressor.c b/examples/regressor.c
index 30021e7f..6dd2b4a2 100644
--- a/examples/regressor.c
+++ b/examples/regressor.c
@@ -1,9 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/option_list.h"
-#include "darknet/blas.h"
-#include "darknet/cuda.h"
+#include "darknet.h"
 #include <sys/time.h>
 #include <assert.h>
 
diff --git a/examples/rnn.c b/examples/rnn.c
index ded23975..0ea4daab 100644
--- a/examples/rnn.c
+++ b/examples/rnn.c
@@ -1,8 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/blas.h"
-#include "darknet/parser.h"
+#include "darknet.h"
 
 typedef struct {
     float *x;
diff --git a/examples/rnn_vid.c b/examples/rnn_vid.c
index a5e2a758..205f24c0 100644
--- a/examples/rnn_vid.c
+++ b/examples/rnn_vid.c
@@ -1,8 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/blas.h"
+#include "darknet.h"
 
 #ifdef OPENCV
 image get_image_from_stream(CvCapture *cap);
diff --git a/examples/segmenter.c b/examples/segmenter.c
index 12eb5408..f24bbdaa 100644
--- a/examples/segmenter.c
+++ b/examples/segmenter.c
@@ -1,9 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/option_list.h"
-#include "darknet/blas.h"
-#include "darknet/cuda.h"
+#include "darknet.h"
 #include <sys/time.h>
 #include <assert.h>
 
@@ -153,7 +148,9 @@ void predict_segmenter(char *datafile, char *cfgfile, char *weightfile, char *fi
         image rgb = mask_to_rgb(m);
         show_image(sized, "orig");
         show_image(rgb, "pred");
+#ifdef OPENCV
         cvWaitKey(0);
+#endif
         printf("Predicted: %f\n", predictions[0]);
         printf("%s: Predicted in %f seconds.\n", input, sec(clock()-time));
         free_image(im);
diff --git a/examples/super.c b/examples/super.c
index d1f1bbf6..89521c94 100644
--- a/examples/super.c
+++ b/examples/super.c
@@ -1,7 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
+#include "darknet.h"
 
 void train_super(char *cfgfile, char *weightfile, int clear)
 {
diff --git a/examples/swag.c b/examples/swag.c
index 91a7beb7..c22d7855 100644
--- a/examples/swag.c
+++ b/examples/swag.c
@@ -1,9 +1,5 @@
-#include "darknet/network.h"
-#include "darknet/detection_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/box.h"
+#include "darknet.h"
+#include <sys/time.h>
 
 void train_swag(char *cfgfile, char *weightfile)
 {
diff --git a/examples/tag.c b/examples/tag.c
index 32a1bd14..f10cea5f 100644
--- a/examples/tag.c
+++ b/examples/tag.c
@@ -1,6 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
+#include "darknet.h"
 
 void train_tag(char *cfgfile, char *weightfile, int clear)
 {
diff --git a/examples/voxel.c b/examples/voxel.c
index b9387bc4..01ea9bb9 100644
--- a/examples/voxel.c
+++ b/examples/voxel.c
@@ -1,7 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
+#include "darknet.h"
 
 void extract_voxel(char *lfile, char *rfile, char *prefix)
 {
diff --git a/examples/writing.c b/examples/writing.c
index 7136ff71..5d30ea90 100644
--- a/examples/writing.c
+++ b/examples/writing.c
@@ -1,6 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
+#include "darknet.h"
 
 void train_writing(char *cfgfile, char *weightfile)
 {
diff --git a/examples/yolo.c b/examples/yolo.c
index c1bef8bd..e6ac873e 100644
--- a/examples/yolo.c
+++ b/examples/yolo.c
@@ -1,10 +1,4 @@
-#include "darknet/network.h"
-#include "darknet/detection_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/box.h"
-#include "darknet/demo.h"
+#include "darknet.h"
 
 char *voc_names[] = {"aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train", "tvmonitor"};
 
diff --git a/include/darknet.h b/include/darknet.h
new file mode 100644
index 00000000..986d4309
--- /dev/null
+++ b/include/darknet.h
@@ -0,0 +1,538 @@
+#ifndef DARKNET_API
+#define DARKNET_API
+#include <stdlib.h>
+
+extern int gpu_index;
+
+#ifdef GPU
+    #define BLOCK 512
+
+    #include "cuda_runtime.h"
+    #include "curand.h"
+    #include "cublas_v2.h"
+
+    #ifdef CUDNN
+    #include "cudnn.h"
+    #endif
+#endif
+
+#ifndef __cplusplus
+    #ifdef OPENCV
+    #include "opencv2/highgui/highgui_c.h"
+    #include "opencv2/imgproc/imgproc_c.h"
+    #include "opencv2/core/version.hpp"
+    #if CV_MAJOR_VERSION == 3
+    #include "opencv2/videoio/videoio_c.h"
+    #endif
+    #endif
+#endif
+
+typedef struct{
+    int *leaf;
+    int n;
+    int *parent;
+    int *child;
+    int *group;
+    char **name;
+
+    int groups;
+    int *group_size;
+    int *group_offset;
+} tree;
+
+typedef enum{
+    LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN
+}ACTIVATION;
+
+typedef enum {
+    CONVOLUTIONAL,
+    DECONVOLUTIONAL,
+    CONNECTED,
+    MAXPOOL,
+    SOFTMAX,
+    DETECTION,
+    DROPOUT,
+    CROP,
+    ROUTE,
+    COST,
+    NORMALIZATION,
+    AVGPOOL,
+    LOCAL,
+    SHORTCUT,
+    ACTIVE,
+    RNN,
+    GRU,
+    CRNN,
+    BATCHNORM,
+    NETWORK,
+    XNOR,
+    REGION,
+    REORG,
+    BLANK
+} LAYER_TYPE;
+
+typedef enum{
+    SSE, MASKED, L1, SMOOTH
+} COST_TYPE;
+
+struct network;
+typedef struct network network;
+
+struct layer;
+typedef struct layer layer;
+
+struct layer{
+    LAYER_TYPE type;
+    ACTIVATION activation;
+    COST_TYPE cost_type;
+    void (*forward)   (struct layer, struct network);
+    void (*backward)  (struct layer, struct network);
+    void (*update)    (struct layer, int, float, float, float);
+    void (*forward_gpu)   (struct layer, struct network);
+    void (*backward_gpu)  (struct layer, struct network);
+    void (*update_gpu)    (struct layer, int, float, float, float);
+    int batch_normalize;
+    int shortcut;
+    int batch;
+    int forced;
+    int flipped;
+    int inputs;
+    int outputs;
+    int nweights;
+    int nbiases;
+    int extra;
+    int truths;
+    int h,w,c;
+    int out_h, out_w, out_c;
+    int n;
+    int max_boxes;
+    int groups;
+    int size;
+    int side;
+    int stride;
+    int reverse;
+    int flatten;
+    int spatial;
+    int pad;
+    int sqrt;
+    int flip;
+    int index;
+    int binary;
+    int xnor;
+    int steps;
+    int hidden;
+    int truth;
+    float smooth;
+    float dot;
+    float angle;
+    float jitter;
+    float saturation;
+    float exposure;
+    float shift;
+    float ratio;
+    float learning_rate_scale;
+    int softmax;
+    int classes;
+    int coords;
+    int background;
+    int rescore;
+    int objectness;
+    int does_cost;
+    int joint;
+    int noadjust;
+    int reorg;
+    int log;
+
+    int adam;
+    float B1;
+    float B2;
+    float eps;
+    int t;
+
+    float alpha;
+    float beta;
+    float kappa;
+
+    float coord_scale;
+    float object_scale;
+    float noobject_scale;
+    float class_scale;
+    int bias_match;
+    int random;
+    float thresh;
+    int classfix;
+    int absolute;
+
+    int onlyforward;
+    int stopbackward;
+    int dontload;
+    int dontloadscales;
+
+    float temperature;
+    float probability;
+    float scale;
+
+    char  * cweights;
+    int   * indexes;
+    int   * input_layers;
+    int   * input_sizes;
+    int   * map;
+    float * rand;
+    float * cost;
+    float * state;
+    float * prev_state;
+    float * forgot_state;
+    float * forgot_delta;
+    float * state_delta;
+
+    float * concat;
+    float * concat_delta;
+
+    float * binary_weights;
+
+    float * biases;
+    float * bias_updates;
+
+    float * scales;
+    float * scale_updates;
+
+    float * weights;
+    float * weight_updates;
+
+    float * delta;
+    float * output;
+    float * squared;
+    float * norms;
+
+    float * spatial_mean;
+    float * mean;
+    float * variance;
+
+    float * mean_delta;
+    float * variance_delta;
+
+    float * rolling_mean;
+    float * rolling_variance;
+
+    float * x;
+    float * x_norm;
+
+    float * m;
+    float * v;
+    
+    float * bias_m;
+    float * bias_v;
+    float * scale_m;
+    float * scale_v;
+
+    float * z_cpu;
+    float * r_cpu;
+    float * h_cpu;
+
+    float * binary_input;
+
+    struct layer *input_layer;
+    struct layer *self_layer;
+    struct layer *output_layer;
+
+    struct layer *input_gate_layer;
+    struct layer *state_gate_layer;
+    struct layer *input_save_layer;
+    struct layer *state_save_layer;
+    struct layer *input_state_layer;
+    struct layer *state_state_layer;
+
+    struct layer *input_z_layer;
+    struct layer *state_z_layer;
+
+    struct layer *input_r_layer;
+    struct layer *state_r_layer;
+
+    struct layer *input_h_layer;
+    struct layer *state_h_layer;
+
+    tree *softmax_tree;
+
+    size_t workspace_size;
+
+    #ifdef GPU
+    int *indexes_gpu;
+
+    float *z_gpu;
+    float *r_gpu;
+    float *h_gpu;
+
+    float *m_gpu;
+    float *v_gpu;
+    float *bias_m_gpu;
+    float *scale_m_gpu;
+    float *bias_v_gpu;
+    float *scale_v_gpu;
+
+    float * prev_state_gpu;
+    float * forgot_state_gpu;
+    float * forgot_delta_gpu;
+    float * state_gpu;
+    float * state_delta_gpu;
+    float * gate_gpu;
+    float * gate_delta_gpu;
+    float * save_gpu;
+    float * save_delta_gpu;
+    float * concat_gpu;
+    float * concat_delta_gpu;
+
+    float *binary_input_gpu;
+    float *binary_weights_gpu;
+
+    float * mean_gpu;
+    float * variance_gpu;
+
+    float * rolling_mean_gpu;
+    float * rolling_variance_gpu;
+
+    float * variance_delta_gpu;
+    float * mean_delta_gpu;
+
+    float * x_gpu;
+    float * x_norm_gpu;
+    float * weights_gpu;
+    float * weight_updates_gpu;
+
+    float * biases_gpu;
+    float * bias_updates_gpu;
+
+    float * scales_gpu;
+    float * scale_updates_gpu;
+
+    float * output_gpu;
+    float * delta_gpu;
+    float * rand_gpu;
+    float * squared_gpu;
+    float * norms_gpu;
+    #ifdef CUDNN
+    cudnnTensorDescriptor_t srcTensorDesc, dstTensorDesc;
+    cudnnTensorDescriptor_t dsrcTensorDesc, ddstTensorDesc;
+    cudnnTensorDescriptor_t normTensorDesc;
+    cudnnFilterDescriptor_t weightDesc;
+    cudnnFilterDescriptor_t dweightDesc;
+    cudnnConvolutionDescriptor_t convDesc;
+    cudnnConvolutionFwdAlgo_t fw_algo;
+    cudnnConvolutionBwdDataAlgo_t bd_algo;
+    cudnnConvolutionBwdFilterAlgo_t bf_algo;
+    #endif
+    #endif
+};
+
+void free_layer(layer);
+
+typedef enum {
+    CONSTANT, STEP, EXP, POLY, STEPS, SIG, RANDOM
+} learning_rate_policy;
+
+typedef struct network{
+    int n;
+    int batch;
+    int *seen;
+    float epoch;
+    int subdivisions;
+    float momentum;
+    float decay;
+    layer *layers;
+    float *output;
+    learning_rate_policy policy;
+
+    float learning_rate;
+    float gamma;
+    float scale;
+    float power;
+    int time_steps;
+    int step;
+    int max_batches;
+    float *scales;
+    int   *steps;
+    int num_steps;
+    int burn_in;
+
+    int adam;
+    float B1;
+    float B2;
+    float eps;
+
+    int inputs;
+    int outputs;
+    int truths;
+    int notruth;
+    int h, w, c;
+    int max_crop;
+    int min_crop;
+    int center;
+    float angle;
+    float aspect;
+    float exposure;
+    float saturation;
+    float hue;
+
+    int gpu_index;
+    tree *hierarchy;
+
+    float *input;
+    float *truth;
+    float *delta;
+    float *workspace;
+    int train;
+    int index;
+    float *cost;
+
+    #ifdef GPU
+    float *input_gpu;
+    float *truth_gpu;
+    float *delta_gpu;
+    float *output_gpu;
+    #endif
+
+} network;
+
+typedef struct {
+    int w;
+    int h;
+    float scale;
+    float rad;
+    float dx;
+    float dy;
+    float aspect;
+} augment_args;
+
+typedef struct {
+    int h;
+    int w;
+    int c;
+    float *data;
+} image;
+
+typedef struct{
+    float x, y, w, h;
+} box;
+
+typedef struct matrix{
+    int rows, cols;
+    float **vals;
+} matrix;
+
+
+typedef struct{
+    int w, h;
+    matrix X;
+    matrix y;
+    int shallow;
+    int *num_boxes;
+    box **boxes;
+} data;
+
+typedef enum {
+    CLASSIFICATION_DATA, DETECTION_DATA, CAPTCHA_DATA, REGION_DATA, IMAGE_DATA, COMPARE_DATA, WRITING_DATA, SWAG_DATA, TAG_DATA, OLD_CLASSIFICATION_DATA, STUDY_DATA, DET_DATA, SUPER_DATA, LETTERBOX_DATA, REGRESSION_DATA, SEGMENTATION_DATA
+} data_type;
+
+typedef struct load_args{
+    int threads;
+    char **paths;
+    char *path;
+    int n;
+    int m;
+    char **labels;
+    int h;
+    int w;
+    int out_w;
+    int out_h;
+    int nh;
+    int nw;
+    int num_boxes;
+    int min, max, size;
+    int classes;
+    int background;
+    int scale;
+    int center;
+    float jitter;
+    float angle;
+    float aspect;
+    float saturation;
+    float exposure;
+    float hue;
+    data *d;
+    image *im;
+    image *resized;
+    data_type type;
+    tree *hierarchy;
+} load_args;
+
+typedef struct{
+    int id;
+    float x,y,w,h;
+    float left, right, top, bottom;
+} box_label;
+
+
+network load_network(char *cfg, char *weights, int clear);
+load_args get_base_args(network net);
+
+void free_data(data d);
+
+typedef struct node{
+    void *val;
+    struct node *next;
+    struct node *prev;
+} node;
+
+typedef struct list{
+    int size;
+    node *front;
+    node *back;
+} list;
+
+pthread_t load_data(load_args args);
+list *read_data_cfg(char *filename);
+list *read_cfg(char *filename);
+
+
+#include "activation_layer.h"
+#include "activations.h"
+#include "avgpool_layer.h"
+#include "batchnorm_layer.h"
+#include "blas.h"
+#include "box.h"
+#include "classifier.h"
+#include "col2im.h"
+#include "connected_layer.h"
+#include "convolutional_layer.h"
+#include "cost_layer.h"
+#include "crnn_layer.h"
+#include "crop_layer.h"
+#include "cuda.h"
+#include "data.h"
+#include "deconvolutional_layer.h"
+#include "demo.h"
+#include "detection_layer.h"
+#include "dropout_layer.h"
+#include "gemm.h"
+#include "gru_layer.h"
+#include "im2col.h"
+#include "image.h"
+#include "layer.h"
+#include "list.h"
+#include "local_layer.h"
+#include "matrix.h"
+#include "maxpool_layer.h"
+#include "network.h"
+#include "normalization_layer.h"
+#include "option_list.h"
+#include "parser.h"
+#include "region_layer.h"
+#include "reorg_layer.h"
+#include "rnn_layer.h"
+#include "route_layer.h"
+#include "shortcut_layer.h"
+#include "softmax_layer.h"
+#include "stb_image.h"
+#include "stb_image_write.h"
+#include "tree.h"
+#include "utils.h"
+#endif
diff --git a/src/activation_kernels.cu b/src/activation_kernels.cu
index f7882990..994e2068 100644
--- a/src/activation_kernels.cu
+++ b/src/activation_kernels.cu
@@ -3,8 +3,8 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/activations.h"
-#include "darknet/cuda.h"
+#include "activations.h"
+#include "cuda.h"
 }
 
 
diff --git a/src/activation_layer.c b/src/activation_layer.c
index 00596f8c..de4e4b7d 100644
--- a/src/activation_layer.c
+++ b/src/activation_layer.c
@@ -1,8 +1,8 @@
-#include "darknet/activation_layer.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
-#include "darknet/gemm.h"
+#include "activation_layer.h"
+#include "utils.h"
+#include "cuda.h"
+#include "blas.h"
+#include "gemm.h"
 
 #include <math.h>
 #include <stdio.h>
diff --git a/src/activations.c b/src/activations.c
index 38a89f2d..0cbb2f55 100644
--- a/src/activations.c
+++ b/src/activations.c
@@ -1,4 +1,4 @@
-#include "darknet/activations.h"
+#include "activations.h"
 
 #include <math.h>
 #include <stdio.h>
diff --git a/src/activations.h b/src/activations.h
index 1c36ff52..54ab3c22 100644
--- a/src/activations.h
+++ b/src/activations.h
@@ -1,12 +1,9 @@
 #ifndef ACTIVATIONS_H
 #define ACTIVATIONS_H
+#include "darknet.h"
 #include "cuda.h"
 #include "math.h"
 
-typedef enum{
-    LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN
-}ACTIVATION;
-
 ACTIVATION get_activation(char *s);
 
 char *get_activation_string(ACTIVATION a);
diff --git a/src/avgpool_layer.c b/src/avgpool_layer.c
index e7158e80..83034dbe 100644
--- a/src/avgpool_layer.c
+++ b/src/avgpool_layer.c
@@ -1,5 +1,5 @@
-#include "darknet/avgpool_layer.h"
-#include "darknet/cuda.h"
+#include "avgpool_layer.h"
+#include "cuda.h"
 #include <stdio.h>
 
 avgpool_layer make_avgpool_layer(int batch, int w, int h, int c)
diff --git a/src/avgpool_layer_kernels.cu b/src/avgpool_layer_kernels.cu
index d8f70adc..a7eca3ae 100644
--- a/src/avgpool_layer_kernels.cu
+++ b/src/avgpool_layer_kernels.cu
@@ -3,8 +3,8 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/avgpool_layer.h"
-#include "darknet/cuda.h"
+#include "avgpool_layer.h"
+#include "cuda.h"
 }
 
 __global__ void forward_avgpool_layer_kernel(int n, int w, int h, int c, float *input, float *output)
diff --git a/src/batchnorm_layer.c b/src/batchnorm_layer.c
index da8ac2fe..4862531d 100644
--- a/src/batchnorm_layer.c
+++ b/src/batchnorm_layer.c
@@ -1,6 +1,6 @@
-#include "darknet/convolutional_layer.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/blas.h"
+#include "convolutional_layer.h"
+#include "batchnorm_layer.h"
+#include "blas.h"
 #include <stdio.h>
 
 layer make_batchnorm_layer(int batch, int w, int h, int c)
diff --git a/src/blas.c b/src/blas.c
index 751c9d7a..668b12c3 100644
--- a/src/blas.c
+++ b/src/blas.c
@@ -1,4 +1,4 @@
-#include "darknet/blas.h"
+#include "blas.h"
 
 #include <math.h>
 #include <assert.h>
diff --git a/src/blas.h b/src/blas.h
index a61b7d8c..49ff0875 100644
--- a/src/blas.h
+++ b/src/blas.h
@@ -18,7 +18,7 @@ void copy_cpu(int N, float *X, int INCX, float *Y, int INCY);
 void scal_cpu(int N, float ALPHA, float *X, int INCX);
 void fill_cpu(int N, float ALPHA, float * X, int INCX);
 float dot_cpu(int N, float *X, int INCX, float *Y, int INCY);
-void test_gpu_blas();
+int test_gpu_blas();
 void shortcut_cpu(int batch, int w1, int h1, int c1, float *add, int w2, int h2, int c2, float *out);
 
 void mean_cpu(float *x, int batch, int filters, int spatial, float *mean);
diff --git a/src/blas_kernels.cu b/src/blas_kernels.cu
index b54603c9..9f1337ca 100644
--- a/src/blas_kernels.cu
+++ b/src/blas_kernels.cu
@@ -4,9 +4,9 @@
 #include <assert.h>
 
 extern "C" {
-#include "darknet/blas.h"
-#include "darknet/cuda.h"
-#include "darknet/utils.h"
+#include "blas.h"
+#include "cuda.h"
+#include "utils.h"
 }
 
 __global__ void scale_bias_kernel(float *output, float *biases, int n, int size)
diff --git a/src/box.c b/src/box.c
index dbf432ee..88ca71ac 100644
--- a/src/box.c
+++ b/src/box.c
@@ -1,4 +1,4 @@
-#include "darknet/box.h"
+#include "box.h"
 #include <stdio.h>
 #include <math.h>
 #include <stdlib.h>
diff --git a/src/box.h b/src/box.h
index 49585ed4..59b38a58 100644
--- a/src/box.h
+++ b/src/box.h
@@ -1,9 +1,6 @@
 #ifndef BOX_H
 #define BOX_H
-
-typedef struct{
-    float x, y, w, h;
-} box;
+#include "darknet.h"
 
 typedef struct{
     float dx, dy, dw, dh;
diff --git a/src/classifier.h b/src/classifier.h
index 3c89f49c..8b137891 100644
--- a/src/classifier.h
+++ b/src/classifier.h
@@ -1,2 +1 @@
 
-list *read_data_cfg(char *filename);
diff --git a/src/col2im_kernels.cu b/src/col2im_kernels.cu
index d8a5e371..aed2df9b 100644
--- a/src/col2im_kernels.cu
+++ b/src/col2im_kernels.cu
@@ -3,8 +3,8 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/col2im.h"
-#include "darknet/cuda.h"
+#include "col2im.h"
+#include "cuda.h"
 }
 
 // src: https://github.com/BVLC/caffe/blob/master/src/caffe/util/im2col.cu
diff --git a/src/compare.c b/src/compare.c
index f773bfbb..4fd266c7 100644
--- a/src/compare.c
+++ b/src/compare.c
@@ -1,11 +1,11 @@
 #include <stdio.h>
 
-#include "darknet/network.h"
-#include "darknet/detection_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/box.h"
+#include "network.h"
+#include "detection_layer.h"
+#include "cost_layer.h"
+#include "utils.h"
+#include "parser.h"
+#include "box.h"
 
 void train_compare(char *cfgfile, char *weightfile)
 {
diff --git a/src/connected_layer.c b/src/connected_layer.c
index 4300dbdc..f8c74235 100644
--- a/src/connected_layer.c
+++ b/src/connected_layer.c
@@ -1,9 +1,9 @@
-#include "darknet/connected_layer.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
-#include "darknet/gemm.h"
+#include "connected_layer.h"
+#include "batchnorm_layer.h"
+#include "utils.h"
+#include "cuda.h"
+#include "blas.h"
+#include "gemm.h"
 
 #include <math.h>
 #include <stdio.h>
diff --git a/src/convolutional_kernels.cu b/src/convolutional_kernels.cu
index 4146a3e5..b53dd16a 100644
--- a/src/convolutional_kernels.cu
+++ b/src/convolutional_kernels.cu
@@ -3,14 +3,14 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/convolutional_layer.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/gemm.h"
-#include "darknet/blas.h"
-#include "darknet/im2col.h"
-#include "darknet/col2im.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
+#include "convolutional_layer.h"
+#include "batchnorm_layer.h"
+#include "gemm.h"
+#include "blas.h"
+#include "im2col.h"
+#include "col2im.h"
+#include "utils.h"
+#include "cuda.h"
 }
 
 __global__ void binarize_kernel(float *x, int n, float *binary)
diff --git a/src/convolutional_layer.c b/src/convolutional_layer.c
index 0dbc7b74..e5b5bb6f 100644
--- a/src/convolutional_layer.c
+++ b/src/convolutional_layer.c
@@ -1,10 +1,10 @@
-#include "darknet/convolutional_layer.h"
-#include "darknet/utils.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/im2col.h"
-#include "darknet/col2im.h"
-#include "darknet/blas.h"
-#include "darknet/gemm.h"
+#include "convolutional_layer.h"
+#include "utils.h"
+#include "batchnorm_layer.h"
+#include "im2col.h"
+#include "col2im.h"
+#include "blas.h"
+#include "gemm.h"
 #include <stdio.h>
 #include <time.h>
 
diff --git a/src/cost_layer.c b/src/cost_layer.c
index 251ec5a9..76c001f0 100644
--- a/src/cost_layer.c
+++ b/src/cost_layer.c
@@ -1,7 +1,7 @@
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
+#include "cost_layer.h"
+#include "utils.h"
+#include "cuda.h"
+#include "blas.h"
 #include <math.h>
 #include <string.h>
 #include <stdlib.h>
diff --git a/src/crnn_layer.c b/src/crnn_layer.c
index a8b040dd..2478fef5 100644
--- a/src/crnn_layer.c
+++ b/src/crnn_layer.c
@@ -1,9 +1,9 @@
-#include "darknet/crnn_layer.h"
-#include "darknet/convolutional_layer.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
-#include "darknet/gemm.h"
+#include "crnn_layer.h"
+#include "convolutional_layer.h"
+#include "utils.h"
+#include "cuda.h"
+#include "blas.h"
+#include "gemm.h"
 
 #include <math.h>
 #include <stdio.h>
diff --git a/src/crop_layer.c b/src/crop_layer.c
index afac7ee0..3b918529 100644
--- a/src/crop_layer.c
+++ b/src/crop_layer.c
@@ -1,5 +1,5 @@
-#include "darknet/crop_layer.h"
-#include "darknet/cuda.h"
+#include "crop_layer.h"
+#include "cuda.h"
 #include <stdio.h>
 
 image get_crop_image(crop_layer l)
diff --git a/src/crop_layer_kernels.cu b/src/crop_layer_kernels.cu
index 41f6f456..b6568219 100644
--- a/src/crop_layer_kernels.cu
+++ b/src/crop_layer_kernels.cu
@@ -3,10 +3,10 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/crop_layer.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
-#include "darknet/image.h"
+#include "crop_layer.h"
+#include "utils.h"
+#include "cuda.h"
+#include "image.h"
 }
 
 __device__ float get_pixel_kernel(float *image, int w, int h, int x, int y, int c)
diff --git a/src/cuda.c b/src/cuda.c
index 085a5764..e011b89d 100644
--- a/src/cuda.c
+++ b/src/cuda.c
@@ -2,9 +2,9 @@ int gpu_index = 0;
 
 #ifdef GPU
 
-#include "darknet/cuda.h"
-#include "darknet/utils.h"
-#include "darknet/blas.h"
+#include "cuda.h"
+#include "utils.h"
+#include "blas.h"
 #include <assert.h>
 #include <stdlib.h>
 #include <time.h>
diff --git a/src/cuda.h b/src/cuda.h
index 52794fa4..cdaba208 100644
--- a/src/cuda.h
+++ b/src/cuda.h
@@ -1,20 +1,10 @@
 #ifndef CUDA_H
 #define CUDA_H
 
-extern int gpu_index;
+#include "darknet.h"
 
 #ifdef GPU
 
-#define BLOCK 512
-
-#include "cuda_runtime.h"
-#include "curand.h"
-#include "cublas_v2.h"
-
-#ifdef CUDNN
-#include "cudnn.h"
-#endif
-
 void check_error(cudaError_t status);
 cublasHandle_t blas_handle();
 float *cuda_make_array(float *x, size_t n);
diff --git a/src/data.c b/src/data.c
index 1ea115a3..dcc5089b 100644
--- a/src/data.c
+++ b/src/data.c
@@ -1,7 +1,7 @@
-#include "darknet/data.h"
-#include "darknet/utils.h"
-#include "darknet/image.h"
-#include "darknet/cuda.h"
+#include "data.h"
+#include "utils.h"
+#include "image.h"
+#include "cuda.h"
 
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/src/data.h b/src/data.h
index c366d66e..8f18d9f6 100644
--- a/src/data.h
+++ b/src/data.h
@@ -2,6 +2,7 @@
 #define DATA_H
 #include <pthread.h>
 
+#include "darknet.h"
 #include "matrix.h"
 #include "list.h"
 #include "image.h"
@@ -17,61 +18,6 @@ static inline float distance_from_edge(int x, int max)
     if (dist > 1) dist = 1;
     return dist;
 }
-
-typedef struct{
-    int w, h;
-    matrix X;
-    matrix y;
-    int shallow;
-    int *num_boxes;
-    box **boxes;
-} data;
-
-typedef enum {
-    CLASSIFICATION_DATA, DETECTION_DATA, CAPTCHA_DATA, REGION_DATA, IMAGE_DATA, COMPARE_DATA, WRITING_DATA, SWAG_DATA, TAG_DATA, OLD_CLASSIFICATION_DATA, STUDY_DATA, DET_DATA, SUPER_DATA, LETTERBOX_DATA, REGRESSION_DATA, SEGMENTATION_DATA
-} data_type;
-
-typedef struct load_args{
-    int threads;
-    char **paths;
-    char *path;
-    int n;
-    int m;
-    char **labels;
-    int h;
-    int w;
-    int out_w;
-    int out_h;
-    int nh;
-    int nw;
-    int num_boxes;
-    int min, max, size;
-    int classes;
-    int background;
-    int scale;
-    int center;
-    float jitter;
-    float angle;
-    float aspect;
-    float saturation;
-    float exposure;
-    float hue;
-    data *d;
-    image *im;
-    image *resized;
-    data_type type;
-    tree *hierarchy;
-} load_args;
-
-typedef struct{
-    int id;
-    float x,y,w,h;
-    float left, right, top, bottom;
-} box_label;
-
-void free_data(data d);
-
-pthread_t load_data(load_args args);
 void load_data_blocking(load_args args);
 
 pthread_t load_data_in_thread(load_args args);
diff --git a/src/deconvolutional_kernels.cu b/src/deconvolutional_kernels.cu
index 1d4bc7c9..16694634 100644
--- a/src/deconvolutional_kernels.cu
+++ b/src/deconvolutional_kernels.cu
@@ -3,15 +3,15 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/convolutional_layer.h"
-#include "darknet/deconvolutional_layer.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/gemm.h"
-#include "darknet/blas.h"
-#include "darknet/im2col.h"
-#include "darknet/col2im.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
+#include "convolutional_layer.h"
+#include "deconvolutional_layer.h"
+#include "batchnorm_layer.h"
+#include "gemm.h"
+#include "blas.h"
+#include "im2col.h"
+#include "col2im.h"
+#include "utils.h"
+#include "cuda.h"
 }
 
 extern "C" void forward_deconvolutional_layer_gpu(layer l, network net)
diff --git a/src/deconvolutional_layer.c b/src/deconvolutional_layer.c
index 3ceff782..3778b418 100644
--- a/src/deconvolutional_layer.c
+++ b/src/deconvolutional_layer.c
@@ -1,11 +1,11 @@
-#include "darknet/deconvolutional_layer.h"
-#include "darknet/convolutional_layer.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/utils.h"
-#include "darknet/im2col.h"
-#include "darknet/col2im.h"
-#include "darknet/blas.h"
-#include "darknet/gemm.h"
+#include "deconvolutional_layer.h"
+#include "convolutional_layer.h"
+#include "batchnorm_layer.h"
+#include "utils.h"
+#include "im2col.h"
+#include "col2im.h"
+#include "blas.h"
+#include "gemm.h"
 
 #include <stdio.h>
 #include <time.h>
diff --git a/src/demo.c b/src/demo.c
index cf22273b..9dc4946a 100644
--- a/src/demo.c
+++ b/src/demo.c
@@ -1,12 +1,12 @@
-#include "darknet/network.h"
-#include "darknet/detection_layer.h"
-#include "darknet/region_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
-#include "darknet/box.h"
-#include "darknet/image.h"
-#include "darknet/demo.h"
+#include "network.h"
+#include "detection_layer.h"
+#include "region_layer.h"
+#include "cost_layer.h"
+#include "utils.h"
+#include "parser.h"
+#include "box.h"
+#include "image.h"
+#include "demo.h"
 #include <sys/time.h>
 
 #define DEMO 1
@@ -31,7 +31,7 @@ static float demo_hier = .5;
 static int running = 0;
 
 static int demo_delay = 0;
-static int demo_frame = 5;
+static int demo_frame = 3;
 static int demo_detections = 0;
 static float **predictions;
 static int demo_index = 0;
diff --git a/src/detection_layer.c b/src/detection_layer.c
index 721d64c2..100694ba 100644
--- a/src/detection_layer.c
+++ b/src/detection_layer.c
@@ -1,10 +1,10 @@
-#include "darknet/detection_layer.h"
-#include "darknet/activations.h"
-#include "darknet/softmax_layer.h"
-#include "darknet/blas.h"
-#include "darknet/box.h"
-#include "darknet/cuda.h"
-#include "darknet/utils.h"
+#include "detection_layer.h"
+#include "activations.h"
+#include "softmax_layer.h"
+#include "blas.h"
+#include "box.h"
+#include "cuda.h"
+#include "utils.h"
 
 #include <stdio.h>
 #include <assert.h>
diff --git a/src/dropout_layer.c b/src/dropout_layer.c
index 1e344f45..780554fb 100644
--- a/src/dropout_layer.c
+++ b/src/dropout_layer.c
@@ -1,6 +1,6 @@
-#include "darknet/dropout_layer.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
+#include "dropout_layer.h"
+#include "utils.h"
+#include "cuda.h"
 #include <stdlib.h>
 #include <stdio.h>
 
diff --git a/src/dropout_layer_kernels.cu b/src/dropout_layer_kernels.cu
index 862e6f45..bd12b678 100644
--- a/src/dropout_layer_kernels.cu
+++ b/src/dropout_layer_kernels.cu
@@ -3,9 +3,9 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/dropout_layer.h"
-#include "darknet/cuda.h"
-#include "darknet/utils.h"
+#include "dropout_layer.h"
+#include "cuda.h"
+#include "utils.h"
 }
 
 __global__ void yoloswag420blazeit360noscope(float *input, int size, float *rand, float prob, float scale)
diff --git a/src/gemm.c b/src/gemm.c
index 245ae641..3003be0e 100644
--- a/src/gemm.c
+++ b/src/gemm.c
@@ -1,6 +1,6 @@
-#include "darknet/gemm.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
+#include "gemm.h"
+#include "utils.h"
+#include "cuda.h"
 #include <stdlib.h>
 #include <stdio.h>
 #include <math.h>
diff --git a/src/gru_layer.c b/src/gru_layer.c
index 68ffed34..7139f798 100644
--- a/src/gru_layer.c
+++ b/src/gru_layer.c
@@ -1,9 +1,9 @@
-#include "darknet/gru_layer.h"
-#include "darknet/connected_layer.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
-#include "darknet/gemm.h"
+#include "gru_layer.h"
+#include "connected_layer.h"
+#include "utils.h"
+#include "cuda.h"
+#include "blas.h"
+#include "gemm.h"
 
 #include <math.h>
 #include <stdio.h>
diff --git a/src/im2col.c b/src/im2col.c
index 67c078bc..69ec98a9 100644
--- a/src/im2col.c
+++ b/src/im2col.c
@@ -1,4 +1,4 @@
-#include "darknet/im2col.h"
+#include "im2col.h"
 #include <stdio.h>
 float im2col_get_pixel(float *im, int height, int width, int channels,
                         int row, int col, int channel, int pad)
diff --git a/src/im2col_kernels.cu b/src/im2col_kernels.cu
index 02f0e070..d42d600b 100644
--- a/src/im2col_kernels.cu
+++ b/src/im2col_kernels.cu
@@ -3,8 +3,8 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/im2col.h"
-#include "darknet/cuda.h"
+#include "im2col.h"
+#include "cuda.h"
 }
 
 // src: https://github.com/BVLC/caffe/blob/master/src/caffe/util/im2col.cu
diff --git a/src/image.c b/src/image.c
index b9f629ca..1a8c984a 100644
--- a/src/image.c
+++ b/src/image.c
@@ -1,14 +1,14 @@
-#include "darknet/image.h"
-#include "darknet/utils.h"
-#include "darknet/blas.h"
-#include "darknet/cuda.h"
+#include "image.h"
+#include "utils.h"
+#include "blas.h"
+#include "cuda.h"
 #include <stdio.h>
 #include <math.h>
 
 #define STB_IMAGE_IMPLEMENTATION
-#include "darknet/stb_image.h"
+#include "stb_image.h"
 #define STB_IMAGE_WRITE_IMPLEMENTATION
-#include "darknet/stb_image_write.h"
+#include "stb_image_write.h"
 
 int windows = 0;
 
diff --git a/src/image.h b/src/image.h
index 7201515c..8b516c14 100644
--- a/src/image.h
+++ b/src/image.h
@@ -7,34 +7,7 @@
 #include <string.h>
 #include <math.h>
 #include "box.h"
-
-#ifndef __cplusplus
-#ifdef OPENCV
-#include "opencv2/highgui/highgui_c.h"
-#include "opencv2/imgproc/imgproc_c.h"
-#include "opencv2/core/version.hpp"
-#if CV_MAJOR_VERSION == 3
-#include "opencv2/videoio/videoio_c.h"
-#endif
-#endif
-#endif
-
-typedef struct {
-    int w;
-    int h;
-    float scale;
-    float rad;
-    float dx;
-    float dy;
-    float aspect;
-} augment_args;
-
-typedef struct {
-    int h;
-    int w;
-    int c;
-    float *data;
-} image;
+#include "darknet.h"
 
 #ifndef __cplusplus
 #ifdef OPENCV
diff --git a/src/layer.c b/src/layer.c
index bfa56e78..c27b4776 100644
--- a/src/layer.c
+++ b/src/layer.c
@@ -1,5 +1,5 @@
-#include "darknet/layer.h"
-#include "darknet/cuda.h"
+#include "layer.h"
+#include "cuda.h"
 
 #include <stdlib.h>
 
diff --git a/src/layer.h b/src/layer.h
index 620d3674..af6cd2ab 100644
--- a/src/layer.h
+++ b/src/layer.h
@@ -1,289 +1 @@
-#ifndef BASE_LAYER_H
-#define BASE_LAYER_H
-
-#include "activations.h"
-#include "stddef.h"
-#include "tree.h"
-
-struct network;
-typedef struct network network;
-
-struct layer;
-typedef struct layer layer;
-
-typedef enum {
-    CONVOLUTIONAL,
-    DECONVOLUTIONAL,
-    CONNECTED,
-    MAXPOOL,
-    SOFTMAX,
-    DETECTION,
-    DROPOUT,
-    CROP,
-    ROUTE,
-    COST,
-    NORMALIZATION,
-    AVGPOOL,
-    LOCAL,
-    SHORTCUT,
-    ACTIVE,
-    RNN,
-    GRU,
-    CRNN,
-    BATCHNORM,
-    NETWORK,
-    XNOR,
-    REGION,
-    REORG,
-    BLANK
-} LAYER_TYPE;
-
-typedef enum{
-    SSE, MASKED, L1, SMOOTH
-} COST_TYPE;
-
-struct layer{
-    LAYER_TYPE type;
-    ACTIVATION activation;
-    COST_TYPE cost_type;
-    void (*forward)   (struct layer, struct network);
-    void (*backward)  (struct layer, struct network);
-    void (*update)    (struct layer, int, float, float, float);
-    void (*forward_gpu)   (struct layer, struct network);
-    void (*backward_gpu)  (struct layer, struct network);
-    void (*update_gpu)    (struct layer, int, float, float, float);
-    int batch_normalize;
-    int shortcut;
-    int batch;
-    int forced;
-    int flipped;
-    int inputs;
-    int outputs;
-    int nweights;
-    int nbiases;
-    int extra;
-    int truths;
-    int h,w,c;
-    int out_h, out_w, out_c;
-    int n;
-    int max_boxes;
-    int groups;
-    int size;
-    int side;
-    int stride;
-    int reverse;
-    int flatten;
-    int spatial;
-    int pad;
-    int sqrt;
-    int flip;
-    int index;
-    int binary;
-    int xnor;
-    int steps;
-    int hidden;
-    int truth;
-    float smooth;
-    float dot;
-    float angle;
-    float jitter;
-    float saturation;
-    float exposure;
-    float shift;
-    float ratio;
-    float learning_rate_scale;
-    int softmax;
-    int classes;
-    int coords;
-    int background;
-    int rescore;
-    int objectness;
-    int does_cost;
-    int joint;
-    int noadjust;
-    int reorg;
-    int log;
-
-    int adam;
-    float B1;
-    float B2;
-    float eps;
-    int t;
-
-    float alpha;
-    float beta;
-    float kappa;
-
-    float coord_scale;
-    float object_scale;
-    float noobject_scale;
-    float class_scale;
-    int bias_match;
-    int random;
-    float thresh;
-    int classfix;
-    int absolute;
-
-    int onlyforward;
-    int stopbackward;
-    int dontload;
-    int dontloadscales;
-
-    float temperature;
-    float probability;
-    float scale;
-
-    char  * cweights;
-    int   * indexes;
-    int   * input_layers;
-    int   * input_sizes;
-    int   * map;
-    float * rand;
-    float * cost;
-    float * state;
-    float * prev_state;
-    float * forgot_state;
-    float * forgot_delta;
-    float * state_delta;
-
-    float * concat;
-    float * concat_delta;
-
-    float * binary_weights;
-
-    float * biases;
-    float * bias_updates;
-
-    float * scales;
-    float * scale_updates;
-
-    float * weights;
-    float * weight_updates;
-
-    float * delta;
-    float * output;
-    float * squared;
-    float * norms;
-
-    float * spatial_mean;
-    float * mean;
-    float * variance;
-
-    float * mean_delta;
-    float * variance_delta;
-
-    float * rolling_mean;
-    float * rolling_variance;
-
-    float * x;
-    float * x_norm;
-
-    float * m;
-    float * v;
-    
-    float * bias_m;
-    float * bias_v;
-    float * scale_m;
-    float * scale_v;
-
-    float * z_cpu;
-    float * r_cpu;
-    float * h_cpu;
-
-    float * binary_input;
-
-    struct layer *input_layer;
-    struct layer *self_layer;
-    struct layer *output_layer;
-
-    struct layer *input_gate_layer;
-    struct layer *state_gate_layer;
-    struct layer *input_save_layer;
-    struct layer *state_save_layer;
-    struct layer *input_state_layer;
-    struct layer *state_state_layer;
-
-    struct layer *input_z_layer;
-    struct layer *state_z_layer;
-
-    struct layer *input_r_layer;
-    struct layer *state_r_layer;
-
-    struct layer *input_h_layer;
-    struct layer *state_h_layer;
-
-    tree *softmax_tree;
-
-    size_t workspace_size;
-
-    #ifdef GPU
-    int *indexes_gpu;
-
-    float *z_gpu;
-    float *r_gpu;
-    float *h_gpu;
-
-    float *m_gpu;
-    float *v_gpu;
-    float *bias_m_gpu;
-    float *scale_m_gpu;
-    float *bias_v_gpu;
-    float *scale_v_gpu;
-
-    float * prev_state_gpu;
-    float * forgot_state_gpu;
-    float * forgot_delta_gpu;
-    float * state_gpu;
-    float * state_delta_gpu;
-    float * gate_gpu;
-    float * gate_delta_gpu;
-    float * save_gpu;
-    float * save_delta_gpu;
-    float * concat_gpu;
-    float * concat_delta_gpu;
-
-    float *binary_input_gpu;
-    float *binary_weights_gpu;
-
-    float * mean_gpu;
-    float * variance_gpu;
-
-    float * rolling_mean_gpu;
-    float * rolling_variance_gpu;
-
-    float * variance_delta_gpu;
-    float * mean_delta_gpu;
-
-    float * x_gpu;
-    float * x_norm_gpu;
-    float * weights_gpu;
-    float * weight_updates_gpu;
-
-    float * biases_gpu;
-    float * bias_updates_gpu;
-
-    float * scales_gpu;
-    float * scale_updates_gpu;
-
-    float * output_gpu;
-    float * delta_gpu;
-    float * rand_gpu;
-    float * squared_gpu;
-    float * norms_gpu;
-    #ifdef CUDNN
-    cudnnTensorDescriptor_t srcTensorDesc, dstTensorDesc;
-    cudnnTensorDescriptor_t dsrcTensorDesc, ddstTensorDesc;
-    cudnnTensorDescriptor_t normTensorDesc;
-    cudnnFilterDescriptor_t weightDesc;
-    cudnnFilterDescriptor_t dweightDesc;
-    cudnnConvolutionDescriptor_t convDesc;
-    cudnnConvolutionFwdAlgo_t fw_algo;
-    cudnnConvolutionBwdDataAlgo_t bd_algo;
-    cudnnConvolutionBwdFilterAlgo_t bf_algo;
-    #endif
-    #endif
-};
-
-void free_layer(layer);
-
-#endif
+#include "darknet.h"
diff --git a/src/list.c b/src/list.c
index 2be48d71..0e4165d3 100644
--- a/src/list.c
+++ b/src/list.c
@@ -1,6 +1,6 @@
 #include <stdlib.h>
 #include <string.h>
-#include "darknet/list.h"
+#include "list.h"
 
 list *make_list()
 {
diff --git a/src/list.h b/src/list.h
index fb818c2a..214b2805 100644
--- a/src/list.h
+++ b/src/list.h
@@ -1,17 +1,6 @@
 #ifndef LIST_H
 #define LIST_H
-
-typedef struct node{
-    void *val;
-    struct node *next;
-    struct node *prev;
-} node;
-
-typedef struct list{
-    int size;
-    node *front;
-    node *back;
-} list;
+#include "darknet.h"
 
 list *make_list();
 int list_find(list *l, void *val);
diff --git a/src/local_layer.c b/src/local_layer.c
index 4fa934c5..aad036e0 100644
--- a/src/local_layer.c
+++ b/src/local_layer.c
@@ -1,9 +1,9 @@
-#include "darknet/local_layer.h"
-#include "darknet/utils.h"
-#include "darknet/im2col.h"
-#include "darknet/col2im.h"
-#include "darknet/blas.h"
-#include "darknet/gemm.h"
+#include "local_layer.h"
+#include "utils.h"
+#include "im2col.h"
+#include "col2im.h"
+#include "blas.h"
+#include "gemm.h"
 #include <stdio.h>
 #include <time.h>
 
diff --git a/src/matrix.c b/src/matrix.c
index 6a0c4fd4..799916bf 100644
--- a/src/matrix.c
+++ b/src/matrix.c
@@ -1,6 +1,6 @@
-#include "darknet/matrix.h"
-#include "darknet/utils.h"
-#include "darknet/blas.h"
+#include "matrix.h"
+#include "utils.h"
+#include "blas.h"
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
diff --git a/src/matrix.h b/src/matrix.h
index d6655eb4..14afcd75 100644
--- a/src/matrix.h
+++ b/src/matrix.h
@@ -1,9 +1,6 @@
 #ifndef MATRIX_H
 #define MATRIX_H
-typedef struct matrix{
-    int rows, cols;
-    float **vals;
-} matrix;
+#include "darknet.h"
 
 matrix make_matrix(int rows, int cols);
 matrix copy_matrix(matrix m);
diff --git a/src/maxpool_layer.c b/src/maxpool_layer.c
index 520dda54..17dedf7a 100644
--- a/src/maxpool_layer.c
+++ b/src/maxpool_layer.c
@@ -1,5 +1,5 @@
-#include "darknet/maxpool_layer.h"
-#include "darknet/cuda.h"
+#include "maxpool_layer.h"
+#include "cuda.h"
 #include <stdio.h>
 
 image get_maxpool_image(maxpool_layer l)
diff --git a/src/maxpool_layer_kernels.cu b/src/maxpool_layer_kernels.cu
index 749c3880..3202e84b 100644
--- a/src/maxpool_layer_kernels.cu
+++ b/src/maxpool_layer_kernels.cu
@@ -3,8 +3,8 @@
 #include "cublas_v2.h"
 
 extern "C" {
-#include "darknet/maxpool_layer.h"
-#include "darknet/cuda.h"
+#include "maxpool_layer.h"
+#include "cuda.h"
 }
 
 __global__ void forward_maxpool_layer_kernel(int n, int in_h, int in_w, int in_c, int stride, int size, int pad, float *input, float *output, int *indexes)
diff --git a/src/network.c b/src/network.c
index 7b753549..2b21338d 100644
--- a/src/network.c
+++ b/src/network.c
@@ -1,34 +1,34 @@
 #include <stdio.h>
 #include <time.h>
 #include <assert.h>
-#include "darknet/network.h"
-#include "darknet/image.h"
-#include "darknet/data.h"
-#include "darknet/utils.h"
-#include "darknet/blas.h"
+#include "network.h"
+#include "image.h"
+#include "data.h"
+#include "utils.h"
+#include "blas.h"
 
-#include "darknet/crop_layer.h"
-#include "darknet/connected_layer.h"
-#include "darknet/gru_layer.h"
-#include "darknet/rnn_layer.h"
-#include "darknet/crnn_layer.h"
-#include "darknet/local_layer.h"
-#include "darknet/convolutional_layer.h"
-#include "darknet/activation_layer.h"
-#include "darknet/detection_layer.h"
-#include "darknet/region_layer.h"
-#include "darknet/normalization_layer.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/maxpool_layer.h"
-#include "darknet/reorg_layer.h"
-#include "darknet/avgpool_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/softmax_layer.h"
-#include "darknet/dropout_layer.h"
-#include "darknet/route_layer.h"
-#include "darknet/shortcut_layer.h"
-#include "darknet/parser.h"
-#include "darknet/data.h"
+#include "crop_layer.h"
+#include "connected_layer.h"
+#include "gru_layer.h"
+#include "rnn_layer.h"
+#include "crnn_layer.h"
+#include "local_layer.h"
+#include "convolutional_layer.h"
+#include "activation_layer.h"
+#include "detection_layer.h"
+#include "region_layer.h"
+#include "normalization_layer.h"
+#include "batchnorm_layer.h"
+#include "maxpool_layer.h"
+#include "reorg_layer.h"
+#include "avgpool_layer.h"
+#include "cost_layer.h"
+#include "softmax_layer.h"
+#include "dropout_layer.h"
+#include "route_layer.h"
+#include "shortcut_layer.h"
+#include "parser.h"
+#include "data.h"
 
 load_args get_base_args(network net)
 {
diff --git a/src/network.h b/src/network.h
index 64d0b61c..15a644e5 100644
--- a/src/network.h
+++ b/src/network.h
@@ -1,81 +1,13 @@
 // Oh boy, why am I about to do this....
 #ifndef NETWORK_H
 #define NETWORK_H
+#include "darknet.h"
 
 #include "image.h"
 #include "layer.h"
 #include "data.h"
 #include "tree.h"
 
-typedef enum {
-    CONSTANT, STEP, EXP, POLY, STEPS, SIG, RANDOM
-} learning_rate_policy;
-
-typedef struct network{
-    int n;
-    int batch;
-    int *seen;
-    float epoch;
-    int subdivisions;
-    float momentum;
-    float decay;
-    layer *layers;
-    float *output;
-    learning_rate_policy policy;
-
-    float learning_rate;
-    float gamma;
-    float scale;
-    float power;
-    int time_steps;
-    int step;
-    int max_batches;
-    float *scales;
-    int   *steps;
-    int num_steps;
-    int burn_in;
-
-    int adam;
-    float B1;
-    float B2;
-    float eps;
-
-    int inputs;
-    int outputs;
-    int truths;
-    int notruth;
-    int h, w, c;
-    int max_crop;
-    int min_crop;
-    int center;
-    float angle;
-    float aspect;
-    float exposure;
-    float saturation;
-    float hue;
-
-    int gpu_index;
-    tree *hierarchy;
-
-
-
-    float *input;
-    float *truth;
-    float *delta;
-    float *workspace;
-    int train;
-    int index;
-    float *cost;
-
-    #ifdef GPU
-    float *input_gpu;
-    float *truth_gpu;
-    float *delta_gpu;
-    float *output_gpu;
-    #endif
-
-} network;
-
 
 #ifdef GPU
 float train_networks(network *nets, int n, data d, int interval);
@@ -118,8 +50,6 @@ void print_network(network net);
 void visualize_network(network net);
 int resize_network(network *net, int w, int h);
 void set_batch_network(network *net, int b);
-network load_network(char *cfg, char *weights, int clear);
-load_args get_base_args(network net);
 void calc_network_cost(network net);
 
 #endif
diff --git a/src/network_kernels.cu b/src/network_kernels.cu
index 56ef50d7..2fbc2dd4 100644
--- a/src/network_kernels.cu
+++ b/src/network_kernels.cu
@@ -7,32 +7,32 @@ extern "C" {
 #include <time.h>
 #include <assert.h>
 
-#include "darknet/network.h"
-#include "darknet/data.h"
-#include "darknet/utils.h"
-#include "darknet/parser.h"
+#include "network.h"
+#include "data.h"
+#include "utils.h"
+#include "parser.h"
 
-#include "darknet/crop_layer.h"
-#include "darknet/connected_layer.h"
-#include "darknet/rnn_layer.h"
-#include "darknet/gru_layer.h"
-#include "darknet/crnn_layer.h"
-#include "darknet/detection_layer.h"
-#include "darknet/region_layer.h"
-#include "darknet/convolutional_layer.h"
-#include "darknet/activation_layer.h"
-#include "darknet/maxpool_layer.h"
-#include "darknet/reorg_layer.h"
-#include "darknet/avgpool_layer.h"
-#include "darknet/normalization_layer.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/local_layer.h"
-#include "darknet/softmax_layer.h"
-#include "darknet/dropout_layer.h"
-#include "darknet/route_layer.h"
-#include "darknet/shortcut_layer.h"
-#include "darknet/blas.h"
+#include "crop_layer.h"
+#include "connected_layer.h"
+#include "rnn_layer.h"
+#include "gru_layer.h"
+#include "crnn_layer.h"
+#include "detection_layer.h"
+#include "region_layer.h"
+#include "convolutional_layer.h"
+#include "activation_layer.h"
+#include "maxpool_layer.h"
+#include "reorg_layer.h"
+#include "avgpool_layer.h"
+#include "normalization_layer.h"
+#include "batchnorm_layer.h"
+#include "cost_layer.h"
+#include "local_layer.h"
+#include "softmax_layer.h"
+#include "dropout_layer.h"
+#include "route_layer.h"
+#include "shortcut_layer.h"
+#include "blas.h"
 }
 
 void forward_network_gpu(network net)
diff --git a/src/normalization_layer.c b/src/normalization_layer.c
index ce62e26e..9f959a4f 100644
--- a/src/normalization_layer.c
+++ b/src/normalization_layer.c
@@ -1,5 +1,5 @@
-#include "darknet/normalization_layer.h"
-#include "darknet/blas.h"
+#include "normalization_layer.h"
+#include "blas.h"
 
 #include <stdio.h>
 
diff --git a/src/option_list.c b/src/option_list.c
index e3cae4d2..f935af30 100644
--- a/src/option_list.c
+++ b/src/option_list.c
@@ -1,8 +1,8 @@
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
-#include "darknet/option_list.h"
-#include "darknet/utils.h"
+#include "option_list.h"
+#include "utils.h"
 
 list *read_data_cfg(char *filename)
 {
diff --git a/src/option_list.h b/src/option_list.h
index 054b3fdd..d0417aa9 100644
--- a/src/option_list.h
+++ b/src/option_list.h
@@ -9,7 +9,6 @@ typedef struct{
 } kvp;
 
 
-list *read_data_cfg(char *filename);
 int read_option(char *s, list *options);
 void option_insert(list *l, char *key, char *val);
 char *option_find(list *l, char *key);
diff --git a/src/parser.c b/src/parser.c
index 38b1ccce..b31e1cd0 100644
--- a/src/parser.c
+++ b/src/parser.c
@@ -3,33 +3,33 @@
 #include <stdlib.h>
 #include <assert.h>
 
-#include "darknet/activation_layer.h"
-#include "darknet/activations.h"
-#include "darknet/avgpool_layer.h"
-#include "darknet/batchnorm_layer.h"
-#include "darknet/blas.h"
-#include "darknet/connected_layer.h"
-#include "darknet/deconvolutional_layer.h"
-#include "darknet/convolutional_layer.h"
-#include "darknet/cost_layer.h"
-#include "darknet/crnn_layer.h"
-#include "darknet/crop_layer.h"
-#include "darknet/detection_layer.h"
-#include "darknet/dropout_layer.h"
-#include "darknet/gru_layer.h"
-#include "darknet/list.h"
-#include "darknet/local_layer.h"
-#include "darknet/maxpool_layer.h"
-#include "darknet/normalization_layer.h"
-#include "darknet/option_list.h"
-#include "darknet/parser.h"
-#include "darknet/region_layer.h"
-#include "darknet/reorg_layer.h"
-#include "darknet/rnn_layer.h"
-#include "darknet/route_layer.h"
-#include "darknet/shortcut_layer.h"
-#include "darknet/softmax_layer.h"
-#include "darknet/utils.h"
+#include "activation_layer.h"
+#include "activations.h"
+#include "avgpool_layer.h"
+#include "batchnorm_layer.h"
+#include "blas.h"
+#include "connected_layer.h"
+#include "deconvolutional_layer.h"
+#include "convolutional_layer.h"
+#include "cost_layer.h"
+#include "crnn_layer.h"
+#include "crop_layer.h"
+#include "detection_layer.h"
+#include "dropout_layer.h"
+#include "gru_layer.h"
+#include "list.h"
+#include "local_layer.h"
+#include "maxpool_layer.h"
+#include "normalization_layer.h"
+#include "option_list.h"
+#include "parser.h"
+#include "region_layer.h"
+#include "reorg_layer.h"
+#include "rnn_layer.h"
+#include "route_layer.h"
+#include "shortcut_layer.h"
+#include "softmax_layer.h"
+#include "utils.h"
 
 typedef struct{
     char *type;
@@ -760,7 +760,7 @@ list *read_cfg(char *filename)
     if(file == 0) file_error(filename);
     char *line;
     int nu = 0;
-    list *sections = make_list();
+    list *options = make_list();
     section *current = 0;
     while((line=fgetl(file)) != 0){
         ++ nu;
@@ -768,7 +768,7 @@ list *read_cfg(char *filename)
         switch(line[0]){
             case '[':
                 current = malloc(sizeof(section));
-                list_insert(sections, current);
+                list_insert(options, current);
                 current->options = make_list();
                 current->type = line;
                 break;
@@ -786,7 +786,7 @@ list *read_cfg(char *filename)
         }
     }
     fclose(file);
-    return sections;
+    return options;
 }
 
 void save_convolutional_weights_binary(layer l, FILE *fp)
diff --git a/src/region_layer.c b/src/region_layer.c
index 49147e35..1f88bf77 100644
--- a/src/region_layer.c
+++ b/src/region_layer.c
@@ -1,9 +1,9 @@
-#include "darknet/region_layer.h"
-#include "darknet/activations.h"
-#include "darknet/blas.h"
-#include "darknet/box.h"
-#include "darknet/cuda.h"
-#include "darknet/utils.h"
+#include "region_layer.h"
+#include "activations.h"
+#include "blas.h"
+#include "box.h"
+#include "cuda.h"
+#include "utils.h"
 
 #include <stdio.h>
 #include <assert.h>
@@ -448,15 +448,65 @@ void forward_region_layer_gpu(const layer l, network net)
         int index = entry_index(l, 0, 0, 5);
         softmax_tree(net.input_gpu + index, l.w*l.h, l.batch*l.n, l.inputs/l.n, 1, l.output_gpu + index, *l.softmax_tree);
         /*
+        // TIMING CODE
+        int zz;
+        int number = 1000;
+        int count = 0;
         int i;
-        int count = 5;
         for (i = 0; i < l.softmax_tree->groups; ++i) {
             int group_size = l.softmax_tree->group_size[i];
-            int index = entry_index(l, 0, 0, count);
-            softmax_gpu(net.input_gpu + index, group_size, l.batch*l.n, l.inputs/l.n, l.w*l.h, 1, l.w*l.h, 1, l.output_gpu + index);
             count += group_size;
         }
+        printf("%d %d\n", l.softmax_tree->groups, count);
+        {
+            double then = what_time_is_it_now();
+            for(zz = 0; zz < number; ++zz){
+                int index = entry_index(l, 0, 0, 5);
+                softmax_tree(net.input_gpu + index, l.w*l.h, l.batch*l.n, l.inputs/l.n, 1, l.output_gpu + index, *l.softmax_tree);
+            }
+            cudaDeviceSynchronize();
+            printf("Good GPU Timing: %f\n", what_time_is_it_now() - then);
+        } 
+        {
+            double then = what_time_is_it_now();
+            for(zz = 0; zz < number; ++zz){
+                int i;
+                int count = 5;
+                for (i = 0; i < l.softmax_tree->groups; ++i) {
+                    int group_size = l.softmax_tree->group_size[i];
+                    int index = entry_index(l, 0, 0, count);
+                    softmax_gpu(net.input_gpu + index, group_size, l.batch*l.n, l.inputs/l.n, l.w*l.h, 1, l.w*l.h, 1, l.output_gpu + index);
+                    count += group_size;
+                }
+            }
+            cudaDeviceSynchronize();
+            printf("Bad GPU Timing: %f\n", what_time_is_it_now() - then);
+        }
+        {
+            double then = what_time_is_it_now();
+            for(zz = 0; zz < number; ++zz){
+                int i;
+                int count = 5;
+                for (i = 0; i < l.softmax_tree->groups; ++i) {
+                    int group_size = l.softmax_tree->group_size[i];
+                    softmax_cpu(net.input + count, group_size, l.batch, l.inputs, l.n*l.w*l.h, 1, l.n*l.w*l.h, l.temperature, l.output + count);
+                    count += group_size;
+                }
+            }
+            cudaDeviceSynchronize();
+            printf("CPU Timing: %f\n", what_time_is_it_now() - then);
+        }
         */
+        /*
+           int i;
+           int count = 5;
+           for (i = 0; i < l.softmax_tree->groups; ++i) {
+           int group_size = l.softmax_tree->group_size[i];
+           int index = entry_index(l, 0, 0, count);
+           softmax_gpu(net.input_gpu + index, group_size, l.batch*l.n, l.inputs/l.n, l.w*l.h, 1, l.w*l.h, 1, l.output_gpu + index);
+           count += group_size;
+           }
+         */
     } else if (l.softmax) {
         int index = entry_index(l, 0, 0, l.coords + !l.background);
         //printf("%d\n", index);
diff --git a/src/reorg_layer.c b/src/reorg_layer.c
index d9f77154..d7fc0010 100644
--- a/src/reorg_layer.c
+++ b/src/reorg_layer.c
@@ -1,6 +1,6 @@
-#include "darknet/reorg_layer.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
+#include "reorg_layer.h"
+#include "cuda.h"
+#include "blas.h"
 
 #include <stdio.h>
 
diff --git a/src/rnn_layer.c b/src/rnn_layer.c
index 998bbacf..fb4f1084 100644
--- a/src/rnn_layer.c
+++ b/src/rnn_layer.c
@@ -1,9 +1,9 @@
-#include "darknet/rnn_layer.h"
-#include "darknet/connected_layer.h"
-#include "darknet/utils.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
-#include "darknet/gemm.h"
+#include "rnn_layer.h"
+#include "connected_layer.h"
+#include "utils.h"
+#include "cuda.h"
+#include "blas.h"
+#include "gemm.h"
 
 #include <math.h>
 #include <stdio.h>
diff --git a/src/route_layer.c b/src/route_layer.c
index 76f80f67..bce66fa4 100644
--- a/src/route_layer.c
+++ b/src/route_layer.c
@@ -1,6 +1,6 @@
-#include "darknet/route_layer.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
+#include "route_layer.h"
+#include "cuda.h"
+#include "blas.h"
 
 #include <stdio.h>
 
diff --git a/src/shortcut_layer.c b/src/shortcut_layer.c
index bf60ddff..b39ba8e8 100644
--- a/src/shortcut_layer.c
+++ b/src/shortcut_layer.c
@@ -1,6 +1,7 @@
-#include "darknet/shortcut_layer.h"
-#include "darknet/cuda.h"
-#include "darknet/blas.h"
+#include "shortcut_layer.h"
+#include "cuda.h"
+#include "blas.h"
+#include "activations.h"
 
 #include <stdio.h>
 #include <assert.h>
diff --git a/src/softmax_layer.c b/src/softmax_layer.c
index c0853f97..7f204802 100644
--- a/src/softmax_layer.c
+++ b/src/softmax_layer.c
@@ -1,6 +1,6 @@
-#include "darknet/softmax_layer.h"
-#include "darknet/blas.h"
-#include "darknet/cuda.h"
+#include "softmax_layer.h"
+#include "blas.h"
+#include "cuda.h"
 
 #include <float.h>
 #include <math.h>
diff --git a/src/tree.c b/src/tree.c
index 89b5a8b4..f36ca3a1 100644
--- a/src/tree.c
+++ b/src/tree.c
@@ -1,8 +1,8 @@
 #include <stdio.h>
 #include <stdlib.h>
-#include "darknet/tree.h"
-#include "darknet/utils.h"
-#include "darknet/data.h"
+#include "tree.h"
+#include "utils.h"
+#include "data.h"
 
 void change_leaves(tree *t, char *leaf_list)
 {
diff --git a/src/tree.h b/src/tree.h
index 2e9c8126..ec74057b 100644
--- a/src/tree.h
+++ b/src/tree.h
@@ -1,18 +1,6 @@
 #ifndef TREE_H
 #define TREE_H
-
-typedef struct{
-    int *leaf;
-    int n;
-    int *parent;
-    int *child;
-    int *group;
-    char **name;
-
-    int groups;
-    int *group_size;
-    int *group_offset;
-} tree;
+#include "darknet.h"
 
 tree *read_tree(char *filename);
 void hierarchy_predictions(float *predictions, int n, tree *hier, int only_leaves, int stride);
diff --git a/src/utils.c b/src/utils.c
index 822fd331..9c06e1a3 100644
--- a/src/utils.c
+++ b/src/utils.c
@@ -6,8 +6,16 @@
 #include <unistd.h>
 #include <float.h>
 #include <limits.h>
+#include <time.h>
 
-#include "darknet/utils.h"
+#include "utils.h"
+
+double what_time_is_it_now()
+{
+    struct timespec now;
+    clock_gettime(CLOCK_REALTIME, &now);
+    return now.tv_sec + now.tv_nsec*1e-9;
+}
 
 int *read_intlist(char *gpu_list, int *ngpus, int d)
 {
@@ -608,13 +616,13 @@ float rand_normal()
 size_t rand_size_t()
 {
     return  ((size_t)(rand()&0xff) << 56) | 
-            ((size_t)(rand()&0xff) << 48) |
-            ((size_t)(rand()&0xff) << 40) |
-            ((size_t)(rand()&0xff) << 32) |
-            ((size_t)(rand()&0xff) << 24) |
-            ((size_t)(rand()&0xff) << 16) |
-            ((size_t)(rand()&0xff) << 8) |
-            ((size_t)(rand()&0xff) << 0);
+        ((size_t)(rand()&0xff) << 48) |
+        ((size_t)(rand()&0xff) << 40) |
+        ((size_t)(rand()&0xff) << 32) |
+        ((size_t)(rand()&0xff) << 24) |
+        ((size_t)(rand()&0xff) << 16) |
+        ((size_t)(rand()&0xff) << 8) |
+        ((size_t)(rand()&0xff) << 0);
 }
 
 float rand_uniform(float min, float max)
diff --git a/src/utils.h b/src/utils.h
index ab4c6959..27206533 100644
--- a/src/utils.h
+++ b/src/utils.h
@@ -7,6 +7,7 @@
 #define SECRET_NUM -1234
 #define TWO_PI 6.2831853071795864769252866
 
+double what_time_is_it_now();
 int *read_intlist(char *s, int *n, int d);
 int *read_map(char *filename);
 void shuffle(void *arr, size_t n, size_t size);