TypeError: Argument 'angle' can not be treated as a double

[Kana-alt]

I'm sorry for asking so many questions.

I'm trying to get to the "train" stage.

I am trying to "CUDA_VISIBLE_DEVICES=0 python train/train_refinenet.py --arch pose_resnet_refine --checkpoint . /checkpoint/all/syn_refinenet --animal all --solver adam --dataset synthetic_animal_sp_all --dataset_real real_animal_all --train_on_all_ cat" was successfully executed.

Then run "CUDA_VISIBLE_DEVICES=0 python train/train_refinenet_mt.py --arch pose_resnet_refine_mt_multida --checkpoint ./checkpoint/all/syn2real_refinenet_mt --dual_branch --mixup_dual --dataset_real_crop real_animal_crop_all --occlusion_aug --pretrained ./checkpoint/all/syn_refinenet_re/model_best.pth.tar --animal all --dataset_real real_animal_all --dataset synthetic_animal_sp_all --train_on_all_cat --generate_pseudol
", the following error occurred.

Please let me know the solution.

Epoch: 1 | LR: 0.000250 | Trg_weight: 15.000000 | C2rcons_weight: 0.606415

==> real_animal_all
Animal:horse, number of frames:13545, train: 8380, valid: 1772
Total number of frames:13545, train: 8380, valid 1772
load from mean file: ./data/synthetic_animal/all_combineds5r5_texture/mean.pth.tar
Real animal mean: 0.3999, 0.3909, 0.3871
Real animal std: 0.2589, 0.2431, 0.2291
Traceback (most recent call last):
File "train/train_refinenet_mt.py", line 547, in
main(parser.parse_args())
File "train/train_refinenet_mt.py", line 182, in main
score_map, generated_kpts = prediction_check(previous_img, previous_kpts, trg_img[i], model, real_dataset_train, device, num_transform=5)
File "C:\Users\kana\Desktop\Synthetic-Animals2\UDA-Animal-Pose-main\train..\CCSSL\scripts\consistency.py", line 59, in prediction_check
M = cv2.getRotationMatrix2D((32,32),-rot,1)
TypeError: Argument 'angle' can not be treated as a double

[chaneyddtt]

Hi @Kana-alt, you are welcome to ask questions. I could not reproduce you error currently. May I know your python version?

[Kana-alt]

I use
Python 3.6.13

[chaneyddtt]

Can you try python 3.7 and opencv-python 4.1.1.26, and see whether the error still occurs?

[Kana-alt]

Thanks for the reply!

I changed the environment and tried again.
Next time I got this error.

==> creating model 'pose_resnet_refine'
Init Network from model zoo
Total params: 29.48M
init all synthetic animal super augmentation
Animal: horse, training samples: 8000, valid samples: 2000
Traceback (most recent call last):
File "train/train_refinenet.py", line 439, in
main(parser.parse_args())
File "train/train_refinenet.py", line 112, in main
train_dataset = datasets.dict[args.dataset](is_train=True, **vars(args))
File "C:\Users\kana\Desktop\UDA-Animal-Pose-main\train..\pose\datasets\synthetic_animal_sp_all.py", line 248, in synthetic_animal_sp_all
return Synthetic_Animal_SP_All(**kwargs)
File "C:\Users\kana\Desktop\UDA-Animal-Pose-main\train..\pose\datasets\synthetic_animal_sp_all.py", line 90, in init
self.load_animal()
File "C:\Users\kana\Desktop\UDA-Animal-Pose-main\train..\pose\datasets\synthetic_animal_sp_all.py", line 128, in load_animal
self.train_set.append(imgpath[train_idxs[idx]])
IndexError: list index out of range

[chaneyddtt]

Do you generate the keypoints_18.json file by yourself? If so, can you check the length of the imgpath by print(len(imgpath)) to check whether you write the data properly?

[Kana-alt]

I've got it figured out! Thank you.