ground-truth generation for ycb_dataset.py

[ll-II]

Hello,

I think there might be a problem with the synthetic data generation: ycb_dataset.py, function gen_synthetic, from line 147:

In my case the class_mask was not correctly generated:

• class_ids was a list of list of class numbers, like: [[3, 16, 20, 6]]
  • cid[0] always 3, leading to several errors

I used this code to generate the keypoint with the function 'gen_kp_gt'

I could fix the issue by changing l. 147 and l. 151 :

class_ids = meta['cls_indexes'][0]
class_mask = np.where(seg_label == cid)

[sjtuytc]

I'm confused, can you provide a visualization to explain why it's wrong?

[sjtuytc]

Feel free to reopen it if necessary.