could you help me this problem?

[SamXiaosheng]

File "/data/sam/Meta-Faster-R-CNN-main/meta_faster_rcnn/modeling/fsod/fsod_fast_rcnn.py", line 112, in fsod_fast_rcnn_inference_single_image
scores = scores.reshape(cls_num, box_num).permute(1, 0)
RuntimeError: shape '[20, 645]' is invalid for input of size 12912

[GuangxingHan]

Can you please share more information about your process, e.g., the log file? which stage are you in?

[SamXiaosheng]

in the first stage , and when testing the few shot. 发自我的iPhone

…

------------------ Original ------------------ From: Guangxing Han ***@***.***> Date: Fri,Dec 9,2022 4:21 AM To: GuangxingHan/Meta-Faster-R-CNN ***@***.***> Cc: Sam ***@***.***>, Author ***@***.***> Subject: Re: [GuangxingHan/Meta-Faster-R-CNN] could you help me this problem?(Issue #21) Can you please share more information about your process, e.g., the log file? which stage are you in? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[GuangxingHan]

Can you please share with me your training log to see what happened there? Thanks.

[SamXiaosheng]

[12/08 17:53:30] d2.data.datasets.coco INFO: Loading datasets/coco/annotations/instances_val2014.json takes 4.73 seconds.
[12/08 17:53:30] d2.data.datasets.coco INFO: Loaded 40504 images in COCO format from datasets/coco/annotations/instances_val2014.json
[12/08 17:53:33] d2.data.build INFO: Distribution of instances among all 80 categories:

category	#instances	category	#instances	category	#instances
person	86348	bicycle	2458	car	14898
motorcycle	3049	airplane	1444	bus	2023
train	1602	truck	3337	boat	3543
traffic light	4351	fire hydrant	650	stop sign	686
parking meter	510	bench	3490	bird	3852
cat	1669	dog	1951	horse	2187
sheep	3119	cow	2788	elephant	1856
bear	462	zebra	1877	giraffe	1767
backpack	2890	umbrella	3919	handbag	4113
tie	2234	suitcase	2219	frisbee	935
skis	2180	snowboard	793	sports ball	2198
kite	2770	baseball bat	1020	baseball gl..	1206
skateboard	1705	surfboard	2227	tennis racket	1626
bottle	8282	wine glass	2620	cup	7010
fork	1775	knife	2555	spoon	2130
bowl	4903	banana	2854	apple	1761
sandwich	1457	orange	2057	broccoli	2679
carrot	2646	hot dog	1010	pizza	2097
donut	2474	cake	2097	chair	12987
couch	1927	potted plant	3069	bed	1450
dining table	5237	toilet	1462	tv	2057
laptop	1781	mouse	850	remote	1863
keyboard	1028	cell phone	2233	microwave	539
oven	1175	toaster	78	sink	1901
refrigerator	888	book	8342	clock	2271
vase	2251	scissors	441	teddy bear	1522
hair drier	74	toothbrush	630
total	288415
[12/08 17:53:33] d2.data.common INFO: Serializing 40504 elements to byte tensors and concatenating them all ...
[12/08 17:53:34] d2.data.common INFO: Serialized dataset takes 153.61 MiB
[12/08 17:53:39] d2.evaluation.evaluator INFO: Start inference on 20252 images
[12/08 17:53:52] d2.engine.train_loop ERROR: Exception during training:
Traceback (most recent call last):
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/detectron2/engine/train_loop.py", line 135, in train

self.after_step()

File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/detectron2/engine/train_loop.py", line 165, in after_step
h.after_step()
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/detectron2/engine/hooks.py", line 353, in after_step
self._do_eval()
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/detectron2/engine/hooks.py", line 328, in _do_eval
results = self._func()
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/detectron2/engine/defaults.py", line 366, in test_and_save_results
self._last_eval_results = self.test(self.cfg, self.model)
File "/data/sam/FCT-main/fsod_mpvit_rcnn_train_net.py", line 197, in test
results_i = inference_on_dataset(model, data_loader, evaluator)
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/detectron2/evaluation/evaluator.py", line 141, in inference_on_dataset
outputs = model(inputs)
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/torch/nn/parallel/distributed.py", line 619, in forward
output = self.module(*inputs[0], **kwargs[0])
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, **kwargs)
File "/data/sam/FCT-main/FCT/modeling/fsod/fsod_mpvit_rcnn.py", line 150, in forward
return self.inference(batched_inputs)
File "/data/sam/FCT-main/FCT/modeling/fsod/fsod_mpvit_rcnn.py", line 471, in inference
results, _ = self.roi_heads.eval_with_support(query_images, query_features_dict, support_proposals_dict, support_box_features_dict)
File "/data/anaconda3/envs/FCT_cu110/lib/python3.7/site-packages/torch/autograd/grad_mode.py", line 26, in decorate_context
return func(*args, **kwargs)
File "/data/sam/FCT-main/FCT/modeling/fsod/fsod_cross_mpvit_roi_heads.py", line 287, in eval_with_support
pred_instances, _ = self.box_predictor.inference(pred_cls, predictions, proposals)
File "/data/sam/FCT-main/FCT/modeling/fsod/fsod_fast_rcnn.py", line 562, in inference
self.test_topk_per_image,
File "/data/sam/FCT-main/FCT/modeling/fsod/fsod_fast_rcnn.py", line 80, in fsod_fast_rcnn_inference
for pred_cls_per_image, scores_per_image, boxes_per_image, image_shape in zip(pred_cls, scores, boxes, image_shapes)
File "/data/sam/FCT-main/FCT/modeling/fsod/fsod_fast_rcnn.py", line 80, in
for pred_cls_per_image, scores_per_image, boxes_per_image, image_shape in zip(pred_cls, scores, boxes, image_shapes)
File "/data/sam/FCT-main/FCT/modeling/fsod/fsod_fast_rcnn.py", line 111, in fsod_fast_rcnn_inference_single_image
scores = scores.reshape(cls_num, box_num).permute(1, 0)
RuntimeError: shape '[20, 653]' is invalid for input of size 13064

[GuangxingHan]

Thanks. But it is hard to see anything here. Have you modified the config file or the related source codes?

Also, could you also print the shape/value of scores, cls_num, box_num to manually check the logic of the codes?

[SamXiaosheng]

i modify cls_num = pred_cls.shape[0]#pred_cls.unique().shape[0] in fsod_fast_rcnn.py file line 108 ,and it can work.