ValueError: operands could not be broadcast together with shapes (133,) (17,)

[deep-practice]

############# Starting Epoch 0 | LR: 0.001 #############
loss: 6.28594531 | acc: 0.6746: 100%|████████████████████████████████████████████████████████████████████████████████| 3122/3122 [16:03<00:00, 3.24it/s]
Train-0 epoch | loss:6.28594531 | acc:0.6746
loading annotations into memory...
Done (t=0.89s)
creating index...
index created!
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 318/318 [00:57<00:00, 5.52it/s]
Traceback (most recent call last):
File "./scripts/train.py", line 267, in
main()
File "./scripts/train.py", line 224, in main
gt_AP = validate_gt(m.module, opt, cfg, heatmap_to_coord)
File "./scripts/train.py", line 172, in validate_gt
res = evaluate_mAP(os.path.join(opt.work_dir, 'test_gt_kpt.json'), ann_type='keypoints', ann_file=os.path.join(cfg.DATASET.VAL.ROOT, cfg.DATASET.VAL.ANN))
File "/home/work/deep_learning/AlphaPose/alphapose/utils/metrics.py", line 99, in evaluate_mAP
cocoEval.evaluate()
File "/home/work/.local/lib/python3.6/site-packages/pycocotools-2.0.0-py3.6-linux-x86_64.egg/pycocotools/cocoeval.py", line 150, in evaluate
for imgId in p.imgIds
File "/home/work/.local/lib/python3.6/site-packages/pycocotools-2.0.0-py3.6-linux-x86_64.egg/pycocotools/cocoeval.py", line 151, in
for catId in catIds}
File "/home/work/.local/lib/python3.6/site-packages/pycocotools-2.0.0-py3.6-linux-x86_64.egg/pycocotools/cocoeval.py", line 223, in computeOks
dx = xd - xg
ValueError: operands could not be broadcast together with shapes (133,) (17,)
Exception ignored in: <alphapose.utils.metrics.evaluate_mAP..NullWriter object at 0x7fc8dd641908>
AttributeError: 'NullWriter' object has no attribute 'flush'

[Fang-Haoshu]

are you trying to train the coco wholebody? you need to change your pycocotools

[deep-practice]

Yes.I'm trainning coco wholebody.
Can you give some hints on how to change pycocotools?

[sid0407]

I am getting the same error. Please tell me how did you correct this error?

[oreo-lp]

I am getting the same error. Please tell me how did you correct this error?

Hi, I have solved this problem, i reinstall numpy == 1.17.4, my cuda == 11.1, and python == 1.8

[Ostyk]

Basically you have to fork the repo and change this line https://github.com/cocodataset/cocoapi/blob/8c9bcc3cf640524c4c20a9c40e89cb6a2f2fa0e9/PythonAPI/pycocotools/cocoeval.py#L523

It seems to be hardcoded to 17 and you cannot pass this as a parameter