kevin-ssy
commented
3 months ago May I know which dataset you are running on?
Closed StuHude closed 3 months ago
kevin-ssy
commented
3 months ago May I know which dataset you are running on?
StuHude
commented
3 months ago cocostuff，After I downloaded the dataset according to your readme and ran evaluate.py, coco.yaml seemed incomplete, after changing ds_name to cocostuff to match evaluta.py code and adding ignore_backgound=False:
The model is running on cuda
5000
Test: [ 0/5000] eta: 15:33:16 mIoU: 0.1095 (0.1095) time: 11.1993 data: 0.0974 max mem: 2309
Test: [ 1/5000] eta: 17:13:29 mIoU: 0.0000 (0.0548) time: 12.4043 data: 0.0495 max mem: 2341
Test: [ 2/5000] eta: 18:58:25 mIoU: 0.0000 (0.0365) time: 13.6666 data: 0.0335 max mem: 2375
Test: [ 3/5000] eta: 18:19:22 mIoU: 0.1178 (0.0568) time: 13.2004 data: 0.0255 max mem: 2375
Test: [ 4/5000] eta: 18:40:27 mIoU: 0.8393 (0.2133) time: 13.4562 data: 0.0207 max mem: 2375
Test: [ 5/5000] eta: 18:32:36 mIoU: 0.0000 (0.1778) time: 13.3646 data: 0.0174 max mem: 2375
Test: [ 6/5000] eta: 18:14:13 mIoU: 0.1357 (0.1718) time: 13.1464 data: 0.0151 max mem: 2375
Test: [ 7/5000] eta: 18:22:25 mIoU: 0.1268 (0.1661) time: 13.2477 data: 0.0134 max mem: 2375
Test: [ 8/5000] eta: 18:40:04 mIoU: 0.6354 (0.2183) time: 13.4625 data: 0.0121 max mem: 2375
Test: [ 9/5000] eta: 18:41:37 mIoU: 0.0000 (0.1965) time: 13.4837 data: 0.0110 max mem: 2375
Test: [ 10/5000] eta: 18:36:35 mIoU: 0.7929 (0.2507) time: 13.4260 data: 0.0101 max mem: 2375
Test: [ 11/5000] eta: 18:42:39 mIoU: 0.3038 (0.2551) time: 13.5016 data: 0.0093 max mem: 2375
Test: [ 12/5000] eta: 18:54:21 mIoU: 0.0000 (0.2355) time: 13.6451 data: 0.0087 max mem: 2375
Traceback (most recent call last):
File "/home/Clip_as_RNN/evaluate.py", line 511, in
and I found that this error is also reported for other data sets, and it is reported during the evaluate phase（VOC：320/1449）
kevin-ssy
commented
3 months ago Thank you for your feedback.
There are several bugs to be fixed:
In the config file config/coco.yaml, you need to add ignore_background: False at line 50 under the parent attribute test
In the file evaluate.py, line 81 and line 264, replace cocostuff to coco
In the file modeling/model/car.py, add the following after line 296:
if len(refined_masks) != len(mask_proposals):
refined_masks = mask_proposalsThank you for pointing this out. We will try to fix that in the original codebase ASAP.
Apology for any inconvenience.
Kevin
Congrats! What fantastic work! But I would like to ask, when I tried to evaluate directly according to your command, I found that it does not run, and coco.yaml seems incomplete, for example, what should its "ignore_background" be set to? And after I set it up, it runs with an error:
Traceback (most recent call last): File "/home/Clip_as_RNN/evaluate.py", line 511, in
main(args)
File "/home/Clip_as_RNN/evaluate.py", line 499, in main
evaluate(
File "/home/Clip_as_RNN/evaluate.py", line 325, in evaluate
label_pred = eval_semantic(
File "/home/Clip_as_RNN/evaluate.py", line 163, in eval_semantic
all_masks, scores = inference_car(
File "/home/Clip_as_RNN/utils/inference_pipeline.py", line 76, in inference_car
masks, scores = car_model(img, sentences)
File "/root/miniconda3/envs/car0/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, *kwargs)
File "/root/miniconda3/envs/car0/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(args, *kwargs)
File "/home/Clip_as_RNN/modeling/model/car.py", line 317, in forward
pseudo_masks, conf_scores = self._forward_car(im_ori, text)
File "/home/Clip_as_RNN/modeling/model/car.py", line 308, in _forward_car
final_predicted_masks = torch.zeros(len(text), refined_masks[0].shape)
IndexError: index 0 is out of bounds for dimension 0 with size 0
It would be very helpful if you could take a look at it. Thank you very much！