NeRF 관련 실험을 3개의 다른 실험환경에서 하고 있는데, 이번에 낸 논문의 실험을 재실험 및 발전시키면서 보니, 성능이 완전히 똑같이 reproduce가 안됐다. Debugging을 해보니 새 실험환경에서의 float32 matrix multiplication할 때 quantization error가 생기고 있었다. Matrix multiplicatoin뿐 아니라 concat operation에서도 생기는 것 같기도. 어느쪽이 문제가 있는 건진 모르겠지만, 혹은 그냥 quantization 방식이 다른 거일수도, 보니까 GTX 1080 ti 와 (RXT 3090, RTX A5000)에서의 feature값들이 달랐다. coordinate transformation등의 계산은 float64로 바꿨다가 다..
https://github.com/rusty1s/pytorch_sparse/issues/182#issuecomment-962950841 Sparse voxel for pointset alongside features · Issue #182 · rusty1s/pytorch_sparse Hello. I have a (B, N, 3) pointsets and (B, N, 1) pointwise features which I wish to voxelize. The goal of this voxelisation is to be able to query from the voxel, (approximated) feature at any giv... github.com 이거면 한 방에 해결될듯? edited 03/05..

https://github.com/hongsukchoi/3DCrowdNet_RELEASE/issues/26 Question about calculation of smpl_trans in Human36M · Issue #26 · hongsukchoi/3DCrowdNet_RELEASE Hi! Firstly, thanks for your excellent work! I was learning your data processing procedure and had confusions about the following code: 3DCrowdNet_RELEASE/data/Human36M/Human36M.py Line 216 in fde8... github.com 예전에 이거 가지고 1주일 고민했던 기억이...

https://github.com/hongsukchoi/generic_tools/blob/master/rendering.py GitHub - hongsukchoi/generic_tools Contribute to hongsukchoi/generic_tools development by creating an account on GitHub. github.com 월드좌표계가 없을 때, 즉 카메라가 한 대만 있을 때, scene 혹은 object의 mean position은 주어졌다고 가정하고 360도 렌더링하는 코드를 짰다. 타겟인 오브젝트의 point가 아니라 neural radiance field를 다른 뷰에서 렌더링하고 싶은지라 카메라를 돌리는 수밖에 없었다. object를 mean position중심..
# affine transform for feature extraction img, trans, mask, bbox = self.affine_transform_and_masking(img, mask, cfg.render_img_shape, mask, expand_ratio=cfg.render_bbox_expand_ratio) K[0, 0] = K[0, 0] * cfg.render_img_shape[1] / bbox[2] K[1, 1] = K[1, 1] * cfg.render_img_shape[0] / bbox[3] K[0, 2] = (K[0, 2] - bbox[0]) * cfg.render_img_shape[1] / bbox[2] K[1, 2] = (K[1, 2] - bbox[1]) * cfg.rende..

https://github.com/microsoft/torchgeo/issues/529 Conda installs a cpu version of pytorch and torchvision · Issue #529 · microsoft/torchgeo Conda installation (tested on Ubuntu 20.04) installs a cpu version of pytorch and torchvision. In my own project, the only remedy I've found was to hardcode the pytorch package to use. Poor sol... github.com 'conda install pytorch' kept throwed me torch.conda..

09.21.2024Open3D 관련https://github.com/isl-org/Open3D/issues/6508 What coordinate system does Open3D use? · Issue #6508 · isl-org/Open3DChecklist I have searched for similar issues. For Python issues, I have tested with the latest development wheel. I have checked the release documentation and the latest documentation (for master b...github.com Why do they make a new camera coordinate system... ..
https://logicatcore.github.io/scratchpad/lidar/sensor-fusion/jupyter/2021/04/20/3D-Oriented-Bounding-Box.html 3D Oriented bounding boxes made simple Calculating 3D oriented bounding boxes. Oriented boxes are useful to avoid obstacles and make best utilitsation of the real navigationable space for autonomous vehicles to steer around. logicatcore.github.io 이거 어떻게 하지 고민했었는데, data point 의 eigen vect..
- Total
- Today
- Yesterday
- part segmentation
- pytorch
- 컴퓨터비전
- pyrender
- densepose
- 머신러닝
- 인터뷰
- VAE
- world coordinate
- focal length
- Generative model
- Pose2Mesh
- 헬스
- nohup
- Machine Learning
- 에디톨로지
- 비전
- Virtual Camera
- Transformation
- 문경식
- 피트니스
- 2d pose
- nerf
- deep learning
- 컴퓨터비젼
- spin
- Docker
- demo
- camera coordinate
- Interview
일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | ||||||
2 | 3 | 4 | 5 | 6 | 7 | 8 |
9 | 10 | 11 | 12 | 13 | 14 | 15 |
16 | 17 | 18 | 19 | 20 | 21 | 22 |
23 | 24 | 25 | 26 | 27 | 28 | 29 |
30 | 31 |