A customized implementation of the paper "StereoNet: guided hierarchical refinement for real-time edge-aware depth prediction"
Customized implementation of the Stereonet guided hierarchical refinement for real-time edge-aware depth prediction
I have just noticed that the share link above is not available now. Please download from this link: https://drive.google.com/file/d/1CO2hGMphCZsNcJPFAk2q4-QG_0zM0LpU/view?usp=sharing.
This project is the first one after I have learned Pytorch. And it's the homework for my CV course. I'm focusing on the segmentation field at present. Thanks for your attention.
optimizer = RMSprop(model.parameters(), lr=1e-3, weight_decay=0.0001)
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.9)
optimizer = optim.Adam(model.parameters(), lr=0.001, betas=(0.9, 0.999))
scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.9)
optimizer = optim.Adam(model.parameters(), lr=0.001, betas=(0.9, 0.999))
if epoch <= 200:
lr = 0.001
else:
lr = 0.0001
optimizer = optim.Adam(model.parameters(), lr=0.001, betas=(0.9, 0.999))
if epoch <= 200:
lr = 0.001
else:
lr = 0.0001
1. wget https://repo.anaconda.com/archive/Anaconda3-5.3.1-Linux-x86_64.sh
2. bash Anaconda3-5.3.1-Linux-x86_64.sh
Please reference to Ubuntu系统下Anaconda使用方法总结 for more information about conda installation.
conda env create -n your_env_name -f environment.yaml
conda activate your_env_name
Pretrain on SceneFlow dataset
cd pretrain-sceneflow
python sceneflow-pretrain.py
Finetune on KITTI 2015
cd finetune-kitti15
python finetune-kitti15.py