Binyr

How to generate 'scale' and 'objpos'?

5

Hi, Thanks for your code! The 'scale' and 'objpos' is also not provided offictially. How do you generate them? Expect for your guidance! Thank you!

Where can I download the extended lip data set?

2

Hi, Thanks for your code! But I just found original LSP dataset with 2000 images in the official website. I am very glad for your guidance!

Is human detection boxes results after NMS?

1

Hi! Thanks for providing such a wonderful work. I have read your code and visualize results.I observe that two almost the same boxes have very different cls_score. I wonder if...

Hi @nihalsid, When I tried to train a model for hypersim ai_001_008 on the data generated by myself, I got a little worse mIoU 0.681 than the model (mIoU: 0.693)...

Hi, This line will sample a torch.float32 sigmas.(https://github.com/pixeli99/SVD_Xtend/blob/609fbf92a3e39a78ed1e2a9e0764a1e912103c48/train_svd.py#L964). As a result, inp_noisy_latents is also converted to torch.float32 (https://github.com/pixeli99/SVD_Xtend/blob/609fbf92a3e39a78ed1e2a9e0764a1e912103c48/train_svd.py#L972). The same problem happens to (https://github.com/pixeli99/SVD_Xtend/blob/609fbf92a3e39a78ed1e2a9e0764a1e912103c48/train_svd.py#L969). Is this torch.float32 type tensor necesaary?...

Hi, thanks for your great work! I download your dataset and find vkitti normal is generated using "discontinuity-aware plane fitting". Does "discontinuity-aware plane fitting" implemented here? https://github.com/baegwangbin/DSINE/blob/main/utils/d2n/plane_svd.py Thank you! Yanrui

Dear auther, Thanks for your great work! The loss weighing here (https://github.com/pixeli99/SVD_Xtend/blob/609fbf92a3e39a78ed1e2a9e0764a1e912103c48/train_svd.py#L1028C17-L1028C25) is different from the weight used in EDM. Can you explain this line for me. Thank you, Yanrui

How to evaluate normal dataset?

1

Hi, thanks for your great work! I wonder how to evaluate your model on normal dataset. It seems there are only scrpts for depth dataset. Thank you, Yanrui

太牛了，这是我见过的最全，也最感性的讲解了，非常想认识博主

1

太牛了，这是我见过的最全，也最感性的讲解了，非常想认识博主