- Start Nvidia-docker : https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#docker
- Checking CUDA, pytorch, python version in paper. (cuda9.1, pytorch 0.4 or 1.0.0 )
- Finding images in dockerHub : https://hub.docker.com/r/iraadit/cuda9.1-cudnn7-opencv-fn
- pull(Download) images into OS.
docker pull iraadit/cuda9.1-cudnn7-opencv-fn:latest
docker images
- Create Container
#보통의 경우
$docker run --name $컨테이너이름 -it (images_id)
#gpu버전의 경우(nvidia-docker)
$docker run --gpus all --name $컨테이너이름 -it (images_id) bash
# Display all of containers
$docker ps -a
- anacodna3, pytorch 1.0.0 install
- Following Gudieline (https://github.com/big-chan/Sejong_computervision_termproject_monodepth)
wget -i splits/kitti_archives_to_download.txt -P kitti_data/
cd kitti_data
unzip "*.zip"
cd ..
CUDA_VISIBLE_DEVICES=0 python train.py --model_name tested --png --log_dir tested_log
- Environment : Ubuntu18.04, TITAN RTX, Docker, CUDA9.1, python3.6.6, pytorch 1.0.1
- Reference DockerHub :
- Trained_model : 모델이 커서 생략합니다.
- Predicted_Submission.npy_link : https://drive.google.com/file/d/1-9192deJaYaqh_VAeHSUw4EOAtBrcaik/view?usp=sharing
- DepthEstimation Task
- Using Docker
- Syntehsized view Task using DeepLearning.
- Various loss
This repo was referenced by https://github.com/big-chan/Sejong_computervision_termproject_monodepth
InClass : https://github.com/sejongresearch/ComputerVision
EvalAIStarters : https://github.com/sejong-rcv/EvalAI-Starters