Kai Xiong

Question about Local refinement unit

2

The paper says " ... to account for the relative importance among the K neighbors, we further regress a spatial weight W ... Then, we modify F_L via a convolution...

Wrong Implementation of the Stochastic Depth ?

1

The authors try to achieve Stochastic Depth by simply set "item.weight.requires_grad" to True/False. However, there are two issues: 1. If a block is drop out during one iteration, all of...

RuntimeError: you can only change requires_grad flags of leaf variables.

1

"item.weight.requires_grad = True" of the forward function of class BlockA, BlockB and BlockC will cause the above error. I found it is caused by the use of DataParallel of the...

In the [script](https://github.com/thodan/bop_toolkit/blob/91d9d7051180ba4976eda321ee51a7ea98d65c3c/scripts/eval_calc_errors.py#L110C72-L110C72), `bool(args.vsd_normalized_by_diameter)` can be replaced by `eval(str(args.vsd_normalized_by_diameter))`, since `bool('True')` and `bool('False')` are both True. The 'str()' operation here handles the case where `args.vsd_normalized_by_diameter==True` and `eval(True)` will break off...

Question about the normalized coordinates and img_aug

2

- Q1: In the data loader func where makes the ground truth kp_uv [link1](https://github.com/rpng/suo_slam/blob/5de01433d177fde5cac4423f05fd554e3c00794e/lib/datasets/bop.py#L654), coordinate (-1,-1) corresponds to the left-bottom bbox corner. However, in the mesh_grid func which is used...

Since the pose predictor outputs `scale=[1,1,1]` by default, we can only adjust the `translation` to make the rendered object has equal size compared to the one in observed image. If...

The code in `renderer/mesh/clip.py` [link](https://github.com/facebookresearch/pytorch3d/blob/89b851e64c7af3a13766462280597a9d06bf9ae7/pytorch3d/renderer/mesh/clip.py#L189) should be modified from ``` if op == ""), (frustum.right, 0, ""), (frustum.bottom, 1, "