
VIBE에 python virtualenv로 환경세팅을 하려그랬는데, 서버 (우분투16.04)에 python3.7이 없었다. softwaree.tistory.com/85 Ubuntu 16.04에서 Python 3.7 설치 들어가며... django를 사용하려면 Python이 최소 3.7 버전이 설치 되어 있어야 합니다. 하지만 우분투 16.04에서 지원하는 버전은 3.5이기 때문에 추가로 설치를 하여야 합니다. 본인의 PC에 설치되어 있 softwaree.tistory.com 일단 이거 참고해서 python3.7 설치하고, virtualenv에서 python3.7을 이용할 수 있도록 github.com/pypa/virtualenv/issues/1059#issuecomment-313867040 Error ..

askubuntu.com/questions/178712/how-to-increase-swap-space How to increase swap space? I have, by default, 250MB of swap space in Ubuntu, and I want to expand it to a larger size. I need 800MB, which I think will be enough to open several applications without having to hit the current askubuntu.com

udpate 2021.03.10 AttributeError: 'GLXPlatform' object has no attribute 'OSMesa' 현재 osmesa를 I2L-MeshNet_RELEASE보고 쓰려 헀는데, 이런 에러가 뜨는데, 알고보니 import pyrender 전에 os.environ["PYOPENGL_PLATFORM"] = "osmesa"를해줘야 되는 것이었다. 그리고 egl이 아니라 osmesa를 쓰게되면 rgb, depth = self.renderer.render(self.scene, flags=render_flags) valid_mask = (depth > 0)[:, :, np.newaxis] output_img = rgb * valid_mask + (1 - valid_mask) ..

2019 ICCV 논문으로 연구실 선배인 경식이형의 논문이다. 이 포스트의 목적은 이 논문의 모듈 중 하나인 'RootNet'에 대해서, 그 중에서도 논문 자체의 contribution보다는, focal length, camera-object distance (depth), per-pixel distance factor에 대해 다룬 내용을 이해하는 것이다. RootNet의 output은 human root joint의 coordinate인 (x,y,Z)이다. x,y는 img coordinate이고, Z는 camera-centered coordinate system 상의 coordiante으로 root joint의 depth다. inference time 때 x,y는 Z를 구한 후에 Z값과 미리 가정한 f..

Update. 2021. 2.19. 1. detectron2 github에서 코드 clone, python path에 densepoe 추가 sys.path.insert(0, '/path/to/detectron2/projects/DensePose/') 2. colab 가이드대로 pyyamml, detectron2를 pip install. 이 때 detectron2의 torch버전이 자기 torch버전과 같도록 주의할 것. 난 1.6을 쓰니까 1.6으로 바꿔줌 3. pip install av 끝. 기본적으로 virtualenv 가상환경에서 https://colab.research.google.com/drive/16jcaJoc6bCFAQ96jDe2HwtXj7BMD_-m5#scrollTo=9_FzH13EjseR..

spin github: https://github.com/nkolot/SPIN spin 깃헙을 보면 demo.py라고 자기들의 mesh output을 image에 projection 시켜서 visualize하는 코드가 있다. mesh를 여전히 3D처럼 보이게 하면서 이미지에 겹쳐보이게 하는 코드로, 3D mesh estimation 연구들에서는 당연히 필요하고 자주 쓰는 기능이다. 나는 내가 짜기 귀찮으니(+어렵) 얘네 것을 갖다 썼다. 그런데 얘네가 pyrender를 통해 mesh를 projection 시키기 위해 camera node를 만들고 camera translation이라는 것을 이 node에 input으로 주는데, 이 부분이 좀 이해가 안 갔다. pred_camera는 얘네 model의 ou..

What is a random walk? markov chain의 특별한 케이스로, 어떤 현상이나 과정을 연속된 랜덤 step으로 설명하는 이론이라고 말할 수 있는 것 같다. https://en.wikipedia.org/wiki/Random_walk 이 영상에서는 random walk를 markov chain이론의 무한개념으로의 확장이라고 설명하는 것 같은데, 사실 여전히 둘의 차이가 명확하지 않다. 기본적으로 markov chain의 정의는 다음과 같다. 한국말로 하면, 한 시점 미래의 state가 오직 현재 state에만 의존하지, 그 전 state들의 history에는 영향받지 않는 현상을 markov chain 이라고 한다. 이 때 한 state에서 다음 state로 넘어가는 관계를 transit..

Densepose data annotation has 24 parts of human body, but their part segmentation masks correspond to 14 parts of the body. You can find some definitions of 24 parts by refering the correspondence listed at the end of this page, https://github.com/facebookresearch/DensePose/blob/master/challenge/2019_COCO_DensePose/data_format.md facebookresearch/DensePose A real-time approach for mapping all hu..

https://github.com/vchoutas/smplify-x/issues/3#issuecomment-507682801 How to fix "Cuda failure src/bvh_cuda_op.cu:877"? · Issue #3 · vchoutas/smplify-x This is a really great work!!! Is anybody get an error when running the script, and how to fix it? (CUDA version:V9.0.176; cudnn version:7.0.5) The error is as follows: Error inside sort: radix_sor... github.com 기본적으로 이 링크대로만 해야 한다. conda로 instal..

이 글에서는 triple product of vectors가 왜 measuring the volume of the parallelepiped(평행육면체)를 의미하는 지 분석한다. 먼저, triple product는 두 개의 vectors를 cross product한 후 나머지 한 개의 vector와 그것을 dot product한다. 수식으로는 다음과 같이 표현된다. 그런데 이게 기하학적으로 보면 다음과 같이 표현할 수 있다. 기하학적으로 triple product가 어떻게 이렇게 되는 지 이해하기 위해서는, cross product를 먼저 볼 필요가 있다. cross product는 마름모의 넓이 곱하기 normal vector라고 할 수 있다. cross product 수식에서 n전까지는 base라고 ..
- Total
- Today
- Yesterday
- part segmentation
- deep learning
- 인터뷰
- pytorch
- spin
- demo
- Machine Learning
- densepose
- 2d pose
- 컴퓨터비전
- 문경식
- VAE
- 피트니스
- nerf
- Transformation
- focal length
- Pose2Mesh
- world coordinate
- pyrender
- 비전
- nohup
- camera coordinate
- 머신러닝
- Generative model
- Virtual Camera
- 컴퓨터비젼
- 헬스
- 에디톨로지
- Interview
- Docker
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 |