Experimental. Use -json_port 8070 to allow connection to the port:8070 to get json-stream

build/darknet/x64/darknet_demo_json_stream.cmd

@@ -0,0 +1,7 @@
+rem Run this file and then open URL in Chrome/Firefox: rem http://localhost:8070
+rem Or open: http://ip-address:8070
+
+darknet.exe detector demo cfg/coco.data cfg/yolov3.cfg yolov3.weights test.mp4 -i 0 -json_port 8070 -dont_show -ext_output
+
+
+pause

build/darknet/x64/darknet_demo_mjpeg_stream.cmd

@@ -1,7 +1,7 @@
 rem Run this file and then open URL in Chrome/Firefox: rem http://localhost:8090
 rem Or open: http://ip-address:8090
 
-darknet.exe detector demo data/voc.data cfg/yolov2-voc.cfg yolo-voc.weights test.mp4 -i 0 -http_port 8090 -dont_show
+darknet.exe detector demo cfg/coco.data cfg/yolov3.cfg yolov3.weights test.mp4 -i 0 -http_port 8090 -dont_show -ext_output
 
 
 pause

include/darknet.h

@@ -744,9 +744,10 @@ LIB_API detection *get_network_boxes(network *net, int w, int h, float thresh, f
 LIB_API void free_detections(detection *dets, int n);
 LIB_API void fuse_conv_batchnorm(network net);
 LIB_API void calculate_binary_weights(network net);
+LIB_API char *detection_to_json(detection *dets, int nboxes, int classes, char **names, long long int frame_id);
 
 LIB_API layer* get_network_layer(network* net, int i);
-LIB_API detection *get_network_boxes(network *net, int w, int h, float thresh, float hier, int *map, int relative, int *num, int letter);
+//LIB_API detection *get_network_boxes(network *net, int w, int h, float thresh, float hier, int *map, int relative, int *num, int letter);
 LIB_API detection *make_network_boxes(network *net, float thresh, int *num);
 LIB_API void reset_rnn(network *net);
 LIB_API float *network_predict_image(network *net, image im);

src/coco.c

@@ -222,7 +222,7 @@ void validate_coco(char *cfgfile, char *weightfile)
             free_image(val_resized[t]);
         }
     }
-    fseek(fp, -2, SEEK_CUR); 
+    fseek(fp, -2, SEEK_CUR);
     fprintf(fp, "\n]\n");
     fclose(fp);
 
@@ -366,6 +366,7 @@ void run_coco(int argc, char **argv)
 {
 	int dont_show = find_arg(argc, argv, "-dont_show");
 	int http_stream_port = find_int_arg(argc, argv, "-http_port", -1);
+    int json_port = find_int_arg(argc, argv, "-json_port", -1);
 	char *out_filename = find_char_arg(argc, argv, "-out_filename", 0);
     char *prefix = find_char_arg(argc, argv, "-prefix", 0);
     float thresh = find_float_arg(argc, argv, "-thresh", .2);
@@ -387,5 +388,5 @@ void run_coco(int argc, char **argv)
     else if(0==strcmp(argv[2], "valid")) validate_coco(cfg, weights);
     else if(0==strcmp(argv[2], "recall")) validate_coco_recall(cfg, weights);
     else if(0==strcmp(argv[2], "demo")) demo(cfg, weights, thresh, hier_thresh, cam_index, filename, coco_classes, 80, frame_skip,
-		prefix, out_filename, http_stream_port, dont_show, ext_output);
+		prefix, out_filename, http_stream_port, json_port, dont_show, ext_output);
 }

src/demo.c

@@ -41,6 +41,8 @@ static int cpp_video_capture = 0;
 static float fps = 0;
 static float demo_thresh = 0;
 static int demo_ext_output = 0;
+static long long int frame_id = 0;
+static int demo_json_port = -1;
 
 static float *predictions[FRAMES];
 static int demo_index = 0;
@@ -114,6 +116,12 @@ void *detect_in_thread(void *ptr)
     det_img = ipl_images[(demo_index + FRAMES / 2 + 1) % FRAMES];
     demo_index = (demo_index + 1)%FRAMES;
 
+    ++frame_id;
+    if (demo_json_port > 0) {
+        int timeout = 200;
+        send_json(dets, nboxes, l.classes, demo_names, frame_id, demo_json_port, timeout);
+    }
+
     draw_detections_cv_v3(det_img, dets, nboxes, demo_thresh, demo_names, demo_alphabet, demo_classes, demo_ext_output);
     free_detections(dets, nboxes);
 
@@ -130,7 +138,7 @@ double get_wall_time()
 }
 
 void demo(char *cfgfile, char *weightfile, float thresh, float hier_thresh, int cam_index, const char *filename, char **names, int classes,
-    int frame_skip, char *prefix, char *out_filename, int http_stream_port, int dont_show, int ext_output)
+    int frame_skip, char *prefix, char *out_filename, int http_stream_port, int json_port, int dont_show, int ext_output)
 {
     //skip = frame_skip;
     image **alphabet = load_alphabet();
@@ -140,6 +148,7 @@ void demo(char *cfgfile, char *weightfile, float thresh, float hier_thresh, int
     demo_classes = classes;
     demo_thresh = thresh;
     demo_ext_output = ext_output;
+    demo_json_port = json_port;
     printf("Demo\n");
     net = parse_network_cfg_custom(cfgfile, 1);    // set batch=1
     if(weightfile){
@@ -355,7 +364,7 @@ void demo(char *cfgfile, char *weightfile, float thresh, float hier_thresh, int
 }
 #else
 void demo(char *cfgfile, char *weightfile, float thresh, float hier_thresh, int cam_index, const char *filename, char **names, int classes,
-    int frame_skip, char *prefix, char *out_filename, int http_stream_port, int dont_show, int ext_output)
+    int frame_skip, char *prefix, char *out_filename, int http_stream_port, int json_port, int dont_show, int ext_output)
 {
     fprintf(stderr, "Demo needs OpenCV for webcam images.\n");
 }

src/demo.h

@@ -2,6 +2,7 @@
 #define DEMO
 
 #include "image.h"
-void demo(char *cfgfile, char *weightfile, float thresh, float hier_thresh, int cam_index, const char *filename, char **names, int classes, int frame_skip, char *prefix, char *out_filename, int http_stream_port, int dont_show, int ext_output);
+void demo(char *cfgfile, char *weightfile, float thresh, float hier_thresh, int cam_index, const char *filename, char **names, int classes,
+    int frame_skip, char *prefix, char *out_filename, int http_stream_port, int json_port, int dont_show, int ext_output);
 
 #endif

src/detector.c

@@ -1352,6 +1352,7 @@ void run_detector(int argc, char **argv)
     int calc_map = find_arg(argc, argv, "-map");
     check_mistakes = find_arg(argc, argv, "-check_mistakes");
     int http_stream_port = find_int_arg(argc, argv, "-http_port", -1);
+    int json_port = find_int_arg(argc, argv, "-json_port", -1);
     char *out_filename = find_char_arg(argc, argv, "-out_filename", 0);
     char *outfile = find_char_arg(argc, argv, "-out", 0);
     char *prefix = find_char_arg(argc, argv, "-prefix", 0);
@@ -1419,7 +1420,7 @@ void run_detector(int argc, char **argv)
             if (strlen(filename) > 0)
                 if (filename[strlen(filename) - 1] == 0x0d) filename[strlen(filename) - 1] = 0;
         demo(cfg, weights, thresh, hier_thresh, cam_index, filename, names, classes, frame_skip, prefix, out_filename,
-            http_stream_port, dont_show, ext_output);
+            http_stream_port, json_port, dont_show, ext_output);
 
         free_list_contents_kvp(options);
         free_list(options);

src/http_stream.cpp

@@ -196,7 +196,8 @@ public:
 };
 // ----------------------------------------
 
-void send_mjpeg(IplImage* ipl, int port, int timeout, int quality) {
+void send_mjpeg(IplImage* ipl, int port, int timeout, int quality)
+{
     static MJPGWriter wri(port, timeout, quality);
     cv::Mat mat = cv::cvarrToMat(ipl);
     wri.write(mat);
@@ -204,6 +205,164 @@ void send_mjpeg(IplImage* ipl, int port, int timeout, int quality) {
 }
 // ----------------------------------------
 
+class JSON_sender
+{
+    SOCKET sock;
+    SOCKET maxfd;
+    fd_set master;
+    int timeout; // master sock timeout, shutdown after timeout millis.
+
+    int _write(int sock, char const*const s, int len)
+    {
+        if (len < 1) { len = strlen(s); }
+        return ::send(sock, s, len, 0);
+    }
+
+public:
+
+    JSON_sender(int port = 0, int _timeout = 200000)
+        : sock(INVALID_SOCKET)
+        , timeout(_timeout)
+    {
+        FD_ZERO(&master);
+        if (port)
+            open(port);
+    }
+
+    ~JSON_sender()
+    {
+        release();
+    }
+
+    bool release()
+    {
+        if (sock != INVALID_SOCKET)
+            ::shutdown(sock, 2);
+        sock = (INVALID_SOCKET);
+        return false;
+    }
+
+    bool open(int port)
+    {
+        sock = ::socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
+
+        SOCKADDR_IN address;
+        address.sin_addr.s_addr = INADDR_ANY;
+        address.sin_family = AF_INET;
+        address.sin_port = htons(port);    // ::htons(port);
+        if (::bind(sock, (SOCKADDR*)&address, sizeof(SOCKADDR_IN)) == SOCKET_ERROR)
+        {
+            cerr << "error : couldn't bind sock " << sock << " to port " << port << "!" << endl;
+            return release();
+        }
+        if (::listen(sock, 10) == SOCKET_ERROR)
+        {
+            cerr << "error : couldn't listen on sock " << sock << " on port " << port << " !" << endl;
+            return release();
+        }
+        FD_ZERO(&master);
+        FD_SET(sock, &master);
+        maxfd = sock;
+        return true;
+    }
+
+    bool isOpened()
+    {
+        return sock != INVALID_SOCKET;
+    }
+
+    bool write(char *outputbuf)
+    {
+        fd_set rread = master;
+        struct timeval to = { 0,timeout };
+        if (::select(maxfd + 1, &rread, NULL, NULL, &to) <= 0)
+            return true; // nothing broken, there's just noone listening
+
+        size_t outlen = strlen(outputbuf);
+
+#ifdef _WIN32
+        for (unsigned i = 0; i<rread.fd_count; i++)
+        {
+            int addrlen = sizeof(SOCKADDR);
+            SOCKET s = rread.fd_array[i];    // fd_set on win is an array, while ...
+#else
+        for (int s = 0; s <= maxfd; s++)
+        {
+            socklen_t addrlen = sizeof(SOCKADDR);
+            if (!FD_ISSET(s, &rread))      // ... on linux it's a bitmask ;)
+                continue;
+#endif
+            if (s == sock) // request on master socket, accept and send main header.
+            {
+                SOCKADDR_IN address = { 0 };
+                SOCKET      client = ::accept(sock, (SOCKADDR*)&address, &addrlen);
+                if (client == SOCKET_ERROR)
+                {
+                    cerr << "error : couldn't accept connection on sock " << sock << " !" << endl;
+                    return false;
+                }
+                maxfd = (maxfd>client ? maxfd : client);
+                FD_SET(client, &master);
+                _write(client, "HTTP/1.0 200 OK\r\n", 0);
+                _write(client,
+                    "Server: Mozarella/2.2\r\n"
+                    "Accept-Range: bytes\r\n"
+                    "Connection: close\r\n"
+                    "Max-Age: 0\r\n"
+                    "Expires: 0\r\n"
+                    "Cache-Control: no-cache, private\r\n"
+                    "Pragma: no-cache\r\n"
+                    "Content-Type: application/json\r\n"
+                    //"Content-Type: multipart/x-mixed-replace; boundary=boundary\r\n"
+                    "\r\n", 0);
+                cerr << "new client " << client << endl;
+            }
+            else // existing client, just stream pix
+            {
+                char head[400];
+                // application/json
+                // application/x-resource+json or application/x-collection+json -  when you are representing REST resources and collections
+                // text/json or text/javascript or text/plain.
+                // https://stackoverflow.com/questions/477816/what-is-the-correct-json-content-type
+                //sprintf(head, "\r\nContent-Length: %zu\r\n\r\n", outlen);
+                //sprintf(head, "--boundary\r\nContent-Type: application/json\r\nContent-Length: %zu\r\n\r\n", outlen);
+                _write(s, head, 0);
+                int n = _write(s, outputbuf, outlen);
+                //cerr << "known client " << s << " " << n << endl;
+                if (n < outlen)
+                {
+                    cerr << "kill client " << s << endl;
+                    ::shutdown(s, 2);
+                    FD_CLR(s, &master);
+                }
+            }
+        }
+        return true;
+    }
+};
+// ----------------------------------------
+
+// JSON format:
+//{
+// "frame_id":8990,
+// "objects":[
+//  {"class_id":4, "name":"aeroplane", "relative coordinates":{"center_x":0.398831, "center_y":0.630203, "width":0.057455, "height":0.020396}, "confidence":0.793070},
+//  {"class_id":14, "name":"bird", "relative coordinates":{"center_x":0.398831, "center_y":0.630203, "width":0.057455, "height":0.020396}, "confidence":0.265497}
+// ]
+//},
+
+void send_json(detection *dets, int nboxes, int classes, char **names, long long int frame_id, int port, int timeout)
+{
+    static JSON_sender js(port, timeout);
+    char *send_buf = detection_to_json(dets, nboxes, classes, names, frame_id);
+
+    js.write(send_buf);
+    std::cout << " JSON-stream sent. \n";
+    free(send_buf);
+}
+
+// ----------------------------------------
+
 CvCapture* get_capture_video_stream(char *path) {
     CvCapture* cap = NULL;
     try {

src/http_stream.h

@@ -15,6 +15,7 @@ extern "C" {
 #include <stdint.h>
 
 #ifdef OPENCV
+void send_json(detection *dets, int nboxes, int classes, char **names, long long int frame_id, int port, int timeout);
 void send_mjpeg(IplImage* ipl, int port, int timeout, int quality);
 CvCapture* get_capture_webcam(int index);
 CvCapture* get_capture_video_stream(char *path);

src/network.c

@@ -658,6 +658,44 @@ void free_detections(detection *dets, int n)
     free(dets);
 }
 
+char *detection_to_json(detection *dets, int nboxes, int classes, char **names, long long int frame_id)
+{
+    const float thresh = 0.005; // function get_network_boxes() has already filtred dets by actual threshold
+
+    char *send_buf = (char *)calloc(1024, sizeof(char));
+    sprintf(send_buf, "{\n \"frame_id\":%d, \n \"objects\":[ \n", frame_id);
+
+    int i, j;
+    int class_id = -1;
+    for (i = 0; i < nboxes; ++i) {
+        for (j = 0; j < classes; ++j) {
+            int show = strncmp(names[j], "dont_show", 9);
+            if (dets[i].prob[j] > thresh && show)
+            {
+                if (class_id != -1) strcat(send_buf, ", \n");
+                class_id = j;
+                char *buf = (char *)calloc(2048, sizeof(char));
+                //sprintf(buf, "{\"image_id\":%d, \"category_id\":%d, \"bbox\":[%f, %f, %f, %f], \"score\":%f}",
+                //    image_id, j, dets[i].bbox.x, dets[i].bbox.y, dets[i].bbox.w, dets[i].bbox.h, dets[i].prob[j]);
+
+                sprintf(buf, "  {\"class_id\":%d, \"name\":\"%s\", \"relative coordinates\":{\"center_x\":%f, \"center_y\":%f, \"width\":%f, \"height\":%f}, \"confidence\":%f}",
+                    j, names[j], dets[i].bbox.x, dets[i].bbox.y, dets[i].bbox.w, dets[i].bbox.h, dets[i].prob[j]);
+
+                int send_buf_len = strlen(send_buf);
+                int buf_len = strlen(buf);
+                int total_len = send_buf_len + buf_len + 100;
+                send_buf = (char *)realloc(send_buf, total_len * sizeof(char));
+                if (!send_buf) return;// exit(-1);
+                strcat(send_buf, buf);
+                free(buf);
+            }
+        }
+    }
+    strcat(send_buf, "\n ] \n}, \n");
+    return send_buf;
+}
+
+
 float *network_predict_image(network *net, image im)
 {
     //image imr = letterbox_image(im, net->w, net->h);

src/yolo.c

@@ -339,6 +339,7 @@ void run_yolo(int argc, char **argv)
 {
 	int dont_show = find_arg(argc, argv, "-dont_show");
 	int http_stream_port = find_int_arg(argc, argv, "-http_port", -1);
+    int json_port = find_int_arg(argc, argv, "-json_port", -1);
 	char *out_filename = find_char_arg(argc, argv, "-out_filename", 0);
     char *prefix = find_char_arg(argc, argv, "-prefix", 0);
     float thresh = find_float_arg(argc, argv, "-thresh", .2);
@@ -359,5 +360,5 @@ void run_yolo(int argc, char **argv)
     else if(0==strcmp(argv[2], "valid")) validate_yolo(cfg, weights);
     else if(0==strcmp(argv[2], "recall")) validate_yolo_recall(cfg, weights);
     else if(0==strcmp(argv[2], "demo")) demo(cfg, weights, thresh, hier_thresh, cam_index, filename, voc_names, 20, frame_skip,
-		prefix, out_filename, http_stream_port, dont_show, ext_output);
+		prefix, out_filename, http_stream_port, json_port, dont_show, ext_output);
 }