-
Hello,
I'm trying to reimplementate your work with A100*8EA.
However, performance was not reached to your one.
I found that some configurations are different compared with your training log.
Could…
-
Hi!
Thanks for your excellent work!
I have some quetions on freihand_train_data.json format.In the training_mano.json annotation file provided with the Freihand dataset, the first 48 denote poses, 4…
-
Hello Team,
When I tried to reproduce this article, I found that my loss value had a jump in about 16 epoch.
Current epoch 15, lr: 0.001, Epoch15 Loss: 0.0290;
Current epoch 16, lr: 0.001, Epoc…
-
Hi, the metrics scores on Freihand Benchmark seems inconsistent between the official arXiv paper and the official repository. Can the authors comment on which is correct? The score reported in the Git…
-
Hi! Thanks for providing such amazing works! I am trying to reproduce your results on the FreiHAND dataset, but the best I obtained is slightly lower than your provided checkpoints. My results for lix…
-
Hi,
I use your pretrained model and test on freihand data as below,
python main/test.py --gpu 0,1,2,3 --cfg ./asset/yaml/pretrained manoJ_test_freihand.yaml
the average error output is 4700.
…
-
Hi,
Thanks for making this awesome project open source. When I try to train the RootNet on the FreiHAND dataset, I fail to find the config for this dataset, such as how to set the `bbox_real`, `pix…
-
Hello, I have got your amazing work, but the test result of the root points on Freihand you give in your 'bbox_root_freihand_output.json' is not as the same as I train the rootnet using the FreiHand D…
-
what is the trainset for hand-only Pose2Pose. Is is right?
## dataset
trainset_3d = ['FreiHAND', 'InterHand26M']
trainset_2d = ['MSCOCO']
testset = 'FreiHAND'
-
Hi, I found that there is no evaluation codes in FreiHAND.py. Could you please provide the codes of FreiHAND to calculate MPJPE and PA-MPJPE or provide some reference links? I think FreiHand is needed…