KeyError: 'max_overlaps'

[Esugi8]

I got ResourceExhaustedError when running train_net.py with original PASCAL VOC 2007, so I reduced the number of Images and Annotations to 1000. Then I tried to run train_net.py again, and got an error below.

What should I do for solving this error? Could you please let me know any solution for this?

Computing bounding-box regression targets... Traceback (most recent call last): File "./faster_rcnn/train_net.py", line 109, in restore=bool(int(args.restore))) File "./faster_rcnn/../lib/fast_rcnn/train.py", line 394, in train_net sw = SolverWrapper(sess, network, imdb, roidb, output_dir, logdir= log_dir, pretrained_model=pretrained_model) File "./faster_rcnn/../lib/fast_rcnn/train.py", line 44, in init self.bbox_means, self.bbox_stds = rdl_roidb.add_bbox_regression_targets(roidb) File "./faster_rcnn/../lib/roi_data_layer/roidb.py", line 65, in add_bbox_regression_targets max_overlaps = roidb[im_i]['max_overlaps'] KeyError: 'max_overlaps'

[CharlesShang]

Hi, @Esugi8 Did you figure out this problem? ResourceExhaustedError sounds like your GPU does not have enough memories. You can try to set TRAIN.SCALES to a small number, like 400.

[liu09114]

did you solve it?

[AuroraLHT]

try clean up the roidb cache file in $TFFRCNN/data/cache. It works for me to solve the 'max_overlaps' problem.