Faster-RCNN训练问题解决
I1013 09:20:29.058217 17752 net.cpp:270] This network produces output cls_prob
I1013 09:20:29.058230 17752 net.cpp:283] Network initialization done.
I1013 09:20:29.122452 17752 net.cpp:816] Ignoring source layer data
I1013 09:20:29.153412 17752 net.cpp:816] Ignoring source layer loss_cls
I1013 09:20:29.153426 17752 net.cpp:816] Ignoring source layer loss_bbox
I1013 09:20:29.153744 17752 net.cpp:816] Ignoring source layer silence_rpn_cls_score
I1013 09:20:29.153748 17752 net.cpp:816] Ignoring source layer silence_rpn_bbox_pred
im_detect: 1/276 0.152s 0.000s
im_detect: 2/276 0.139s 0.000s
im_detect: 3/276 0.134s 0.000s
im_detect: 4/276 0.132s 0.000s
im_detect: 5/276 0.130s 0.000sim_detect: 272/276 0.124s 0.000s
im_detect: 273/276 0.124s 0.000s
im_detect: 274/276 0.124s 0.000s
im_detect: 275/276 0.124s 0.000s
im_detect: 276/276 0.124s 0.000s
Evaluating detections
Writing obj VOC results file
VOC07 metric? Yes
Traceback (most recent call last):
File "./tools/test_net.py", line 90, in
test_net(net, imdb, max_per_image=args.max_per_image, vis=args.vis)
File "/home/py-faster-rcnn/tools/../lib/fast_rcnn/test.py", line 295, in test_net
imdb.evaluate_detections(all_boxes, output_dir)
File "/home/py-faster-rcnn/tools/../lib/datasets/pascal_voc.py", line 332, in evaluate_detections
self._do_python_eval(output_dir)
File "/home/py-faster-rcnn/tools/../lib/datasets/pascal_voc.py", line 295, in _do_python_eval
use_07_metric=use_07_metric)
File "/home/py-faster-rcnn/tools/../lib/datasets/voc_eval.py", line 126, in voc_eval
R = [obj for obj in recs[imagename] if obj['name'] == classname]
KeyError: 'IMG_0805'
解决: 删除data/VOCdekit2007下的annotations_cache文件夹
on.
I1014 05:35:49.888906 31318 net.cpp:228] conv3 does not need backward computation.
I1014 05:35:49.888907 31318 net.cpp:228] pool2 does not need backward computation.
I1014 05:35:49.888909 31318 net.cpp:228] norm2 does not need backward computation.
I1014 05:35:49.888911 31318 net.cpp:228] relu2 does not need backward computation.
I1014 05:35:49.888912 31318 net.cpp:228] conv2 does not need backward computation.
I1014 05:35:49.888914 31318 net.cpp:228] pool1 does not need backward computation.
I1014 05:35:49.888916 31318 net.cpp:228] norm1 does not need backward computation.
I1014 05:35:49.888918 31318 net.cpp:228] relu1 does not need backward computation.
I1014 05:35:49.888919 31318 net.cpp:228] conv1 does not need backward computation.
I1014 05:35:49.888921 31318 net.cpp:270] This network produces output bbox_pred
I1014 05:35:49.888922 31318 net.cpp:270] This network produces output cls_prob
I1014 05:35:49.888936 31318 net.cpp:283] Network initialization done.
I1014 05:35:49.957736 31318 net.cpp:816] Ignoring source layer data
I1014 05:35:49.988982 31318 net.cpp:816] Ignoring source layer loss_cls
I1014 05:35:49.988991 31318 net.cpp:816] Ignoring source layer loss_bbox
I1014 05:35:49.989269 31318 net.cpp:816] Ignoring source layer silence_rpn_cls_score
I1014 05:35:49.989274 31318 net.cpp:816] Ignoring source layer silence_rpn_bbox_pred
im_detect: 1/276 0.155s 0.000s
im_detect: 2/276 0.139s 0.000s
im_detect: 3/276 0.133s 0.000s
im_detect: 4/276 0.130s 0.000s
im_detect: 5/276 0.129s 0.000s
im_detect: 6/276 0.127s 0.000s
im_detect: 7/276 0.127s 0.000s
im_detect: 8/276 0.127s 0.000s
im_detect: 9/276 0.126s 0.000s
im_detect: 10/276 0.126s 0.000s
im_detect: 11/276 0.125s 0.000s
im_detect: 12/276 0.125s 0.000s
im_detect: 13/276 0.125s 0.000s
im_detect: 14/276 0.125s 0.000s
im_detect: 15/276 0.125s 0.000s
im_detect: 16/276 0.125s 0.000s
im_detect: 17/276 0.125s 0.000s
im_detect: 18/276 0.124s 0.000s
im_detect: 19/276 0.124s 0.000s
im_detect: 20/276 0.124s 0.000s
im_detect: 21/276 0.124s 0.000s
im_detect: 22/276 0.124s 0.000s
im_detect: 263/276 0.122s 0.000s
im_detect: 264/276 0.122s 0.000s
im_detect: 265/276 0.122s 0.000s
im_detect: 266/276 0.122s 0.000s
im_detect: 267/276 0.122s 0.000s
im_detect: 268/276 0.122s 0.000s
im_detect: 269/276 0.122s 0.000s
im_detect: 270/276 0.122s 0.000s
im_detect: 271/276 0.122s 0.000s
im_detect: 272/276 0.122s 0.000s
im_detect: 273/276 0.122s 0.000s
im_detect: 274/276 0.122s 0.000s
im_detect: 275/276 0.122s 0.000s
im_detect: 276/276 0.122s 0.000s
Evaluating detections
Writing obj VOC results file
VOC07 metric? Yes
Reading annotation for 1/276
Reading annotation for 101/276
Reading annotation for 201/276
Saving cached annotations to /home/py-faster-rcnn/data/VOCdevkit2007/annotations_cache/annots.pkl
AP for obj = 0.1282
Mean AP = 0.1282
~~~~~~~~
Results:
0.128
0.128
~~~~~~~~
--------------------------------------------------------------
Results computed with the **unofficial** Python eval code.
Results should be very close to the official MATLAB eval code.
Recompute with `./tools/reval.py --matlab ...` for your paper.
-- Thanks, The Management
--------------------------------------------------------------
real 0m59.349s
user 0m59.772s
sys 0m3.548s