how do you generate 'smplx_joints_2d'

[liwenssss]

Hi, thanks for your nice work. I notice that there are some keys such as 'smplx_joints_2d_path', 'smplx_joints_3d_path' in your annotation files. I have a question about how do you generate the relative files.

[mks0601]

Hi

Sorry for confusion. I updated how to get AGORA annotation files at here. Please let me know if you meet any problems.

[liwenssss]

Hi, another question. I download AGORA image in 1280x720 and I directly run the dataset of AGORA('train' split) in order to see the cropped image and the bbox of hand. But it seems that there is something wrong .. Should I need extra preprocessing?

[mks0601]

Could you explain more? What do you mean by running the dataset of AGORA('train' split)? How did you visualize the hand boxes?

[liwenssss]

Yes, I just use your debug code:

            if lhand_bbox_valid:
                _tmp = lhand_bbox.copy().reshape(2, 2)
                _tmp[:, 0] = _tmp[:, 0] / cfg.output_hm_shape[2] * cfg.input_img_shape[1]
                _tmp[:, 1] = _tmp[:, 1] / cfg.output_hm_shape[1] * cfg.input_img_shape[0]
                cv2.rectangle(_img, (int(_tmp[0, 0]), int(_tmp[0, 1])), (int(_tmp[1, 0]), int(_tmp[1, 1])), (255, 0, 0),
                              3)
                cv2.imshow('img', _img)
                cv2.waitKey(0)

another settings:

class AGORA(torch.utils.data.Dataset):
    def __init__(self, transform, data_split):
        self.transform = transform
        self.data_split = data_split
        self.data_path = '/home/lws/dataset/AGORA/'
        self.resolution = (720, 1280)  # height, width. one of (720, 1280) and (2160, 3840)
        self.test_set = 'val'  # val, test

from torch.utils.data.dataloader import DataLoader
    dataset = AGORA(torchvision.transforms.ToTensor(), 'train')
    data_loader = DataLoader(dataset, batch_size=1, shuffle=False)

    for _ in data_loader:
        pass

[mks0601]

Sorry, give me a second

[mks0601]

I fixed all bugs. Thanks!

[liwenssss]

Hi, thanks for you reply again! Maybe some little error, I am not sure:

1. https://github.com/mks0601/Hand4Whole_RELEASE/blob/62bb511de0e1957394d312fd33372ea2f36a267b/tool/AGORA/agora2coco.py#L50
   dataset_path shoule be root_path?
2. https://github.com/mks0601/Hand4Whole_RELEASE/blob/62bb511de0e1957394d312fd33372ea2f36a267b/tool/AGORA/agora2coco.py#L99 Probably no need to add the second item

[mks0601]

1. You can see the final directory in README.
2. The code processes both SMPL and SMPLX. There is a SMPL-only protocol in AGORA.