jaehakim97 / blurhand_release Goto Github PK

Hi, I followed your process to set up the environment but encountered an error when using the pre-trained model for inference. It seems to be an issue with the environment. Could you offer me some advice? I would be very grateful.

I'm working on a similar project and would greatly appreciate it if you could share the implementation code for the model's inference predictions about a single hand picture.

I've inferred several images and found that the gestures are quite similar, but the rotation angle of the wrist joint is not consistent with the images. I would like to ask whether the predicted 21 keypoints can be mapped to the real coordinates of the images in the model design.

Can you provide a demo for inferring local images? I am currently unable to obtain the 21 2D joint coordinate at the image level. Thank you~

hello,tks for this amazing work. And how to run this on my own dataset without annot_file?

Dear,
While I have followed the instructions provided in README, I found that the images of testset are fully absent and those of trainset are also partially absent, which keeps raising errors when running the training script below:

CUDA_VISIBLE_DEVICES=0,1 python src/train.py -opt options/train/BlurHandNet_BH.yml

Note that annotation and image data are located well at the path where BlurHandNet_BH.yml specifies

I respectfully ask whether the missing materials are being ready for the release or I have missed something (if so, I am sorry).

By the way, thanks for your great work!