Commit Graph

892 Commits

Author SHA1 Message Date
e96401774b Fixed compile error 2018-05-08 19:09:43 +03:00
2c2ba0f8fa Update Readme.md 2018-05-08 18:08:21 +03:00
fb56f6d569 Fixed README.md 2018-05-08 18:02:23 +03:00
0948df52b8 Merge pull request #741 from IlyaOvodov/Fix_detector_output
Output improvements for detector results:
2018-05-08 17:27:45 +03:00
1b5afb4583 Output improvements for detector results:
When printing detector results, output was done in random order, obfuscating results for interpreting. Now:
1. Text output includes coordinates of rects in (left,right,top,bottom in pixels) along with label and score
2. Text output is sorted by rect lefts to simplify finding appropriate rects on image
3. If several class probs are > thresh for some detection, the most probable is written first and coordinates for others are not repeated
4. Rects are imprinted in image in order by their best class prob, so most probable rects are always on top and not overlayed by less probable ones
5. Most probable label for rect is always written first
Also:
6. Message about low GPU memory include required amount

Upd:
* struct detection changes reverted to keep it unchanged
* cmd line key "-ext_output" is added to manage extended output (rect coords in detector test mode)

# Conflicts:
#	src/detector.c
2018-05-08 14:09:35 +03:00
0045074f56 Get pre-trained weights for Yolo v3 tiny 2018-05-07 23:33:10 +03:00
ae4ec1a06d Exit with nonzero status on error 2018-05-07 22:11:57 +02:00
6b8fd6f33f Update Readme.md 2018-05-07 02:18:06 +03:00
bde2a30603 Merge pull request #760 from PuneetKohli/fix-log-parser-no-display
Change matplotlib backend to 'agg' when 'show_plot' is false.
2018-05-07 01:36:45 +03:00
cdd1cb0e8c Change matplotlib backend to 'agg' when 'show_plot' is false. This allows the script to be run from command line without any errors (for ex, on a remote machine) 2018-05-06 16:45:26 -05:00
c9b8bdee18 Minor fix - what pip-packages are required for Python scripts 2018-05-06 21:51:31 +03:00
30e08573d1 Merge pull request #759 from Adesun/log_parser
darknet training log parser
2018-05-06 21:31:49 +03:00
efcbdb5425 Added yolov3-tiny.cfg 2018-05-06 21:20:01 +03:00
Ade
a1c5bae6fb Update readme.md 2018-05-06 22:16:31 +08:00
d3d4c0e699 parsing training log and plot
save loss data to csv
2018-05-06 22:11:49 +08:00
85388d6708 Merge pull request #754 from tigerhawkvok/python3-detection-runner
Python3 detection runner
2018-05-05 15:46:49 +03:00
51fbf01277 Address GPU message in root dir 2018-05-04 15:07:45 -07:00
75295718e1 Address GPU message in build dir 2018-05-04 15:07:28 -07:00
1aadf29a51 fix detection paths 2018-05-04 15:00:25 -07:00
3a2a8653a6 Fix detection paths 2018-05-04 14:59:29 -07:00
29f3c37314 Remove debugging line 2018-05-04 14:47:04 -07:00
9c400dc5b8 Remove debugging line 2018-05-04 14:46:33 -07:00
3459c47bfa Also replace root version to Python3 2018-05-04 14:00:11 -07:00
eaecb49a5f Change darknet.py to Python 3 compatibility 2018-05-04 13:59:28 -07:00
89354d0a0c Fixed memory leaks. And fixes for Web-camera and IP-camera. 2018-05-04 23:52:05 +03:00
4232ace376 Fix webcam memory leak 2018-05-04 18:37:33 +03:00
028696bf15 Output improvements for detector results:
When printing detector results, output was done in random order, obfuscating results for interpreting. Now:
1. Text output includes coordinates of rects in (left,right,top,bottom in pixels) along with label and score
2. Text output is sorted by rect lefts to simplify finding appropriate rects on image
3. If several class probs are > thresh for some detection, the most probable is written first and coordinates for others are not repeated
4. Rects are imprinted in image in order by their best class prob, so most probable rects are always on top and not overlayed by less probable ones
5. Most probable label for rect is always written first
Also:
6. Message about low GPU memory include required amount
2018-05-03 16:33:46 +03:00
535353cdaa Merge pull request #737 from IlyaOvodov/Fix_thin
FIX #736: Detector does not draw rects for images with small resolutions
2018-05-03 13:18:29 +03:00
92b25a7200 FIX #736: Detector does not draw rects for images with small resolutions 2018-05-03 12:43:48 +03:00
a0e288f0dc Increased speed of darknet.py. Fixed functions related to resize_image(). 2018-05-01 02:17:49 +03:00
5e3dcb6f34 Update Readme.md 2018-04-26 17:01:48 +03:00
15139eb6ab Fixed TP/FP/FN, IoU calculation for specified threshold 2018-04-24 22:26:55 +03:00
3df335bb50 Fixed SSE4.1 dependencies when AVX=1 on Linux 2018-04-23 23:15:21 +03:00
a2f3644e87 Merge pull request #675 from IlyaOvodov/Flip_Fix
Fix of issue #674: parameter "flip" is added to [net] parameters for cla…
2018-04-21 23:56:46 +03:00
9207607a59 Fix issue #674: parameter "flip" is added to [net] parameters for classifier and detector
flip=0 switches flip augmentation off.
Default is flip=1 which is consistent with a previous behaviour.
2018-04-21 23:22:16 +03:00
e18a303d1c Fixed utils.c for short lines. 2018-04-21 15:19:58 +03:00
5d616450a4 Usage of calc_anchors 2018-04-20 02:04:02 +03:00
5767893292 Fixed typo for calc_anchors 2018-04-20 01:59:04 +03:00
c2c8595b08 Fixed detection and validate(recall, map) for .png and .bmp images. 2018-04-19 14:47:39 +03:00
eccfccdaf7 Focal loss fixed 2018-04-19 01:56:29 +03:00
6056b835eb Fixed Focal-loss 2018-04-19 01:34:26 +03:00
57fce97fd5 yolo_console_dll.cpp uses Yolo v3 by default 2018-04-18 22:17:24 +03:00
8fcf3f04ee sure ODR 2018-04-18 16:00:06 +03:00
9f2dbfa7df Added flag AVX to the Makefile 2018-04-17 14:39:27 +03:00
6e3afec85a Fixed Python detection 2018-04-17 14:21:23 +03:00
140333977c Update Readme.md 2018-04-17 03:06:46 +03:00
9bae70b225 Accelerated by another 5% using FP16/32 Batch-norm for Tensor Cores. 2018-04-17 02:51:11 +03:00
701f4fab63 Compile fix 2018-04-16 17:58:08 +03:00
892137f3aa Added CUDNN_HALF to the Makefile 2018-04-16 17:52:03 +03:00
823c3f4b28 Fixed AVX detection on Linux. 2018-04-16 16:01:35 +03:00