Skip to content
/ IFR Public

[CVPR'22] Semi-Supervised Video Semantic Segmentation with Inter-Frame Feature Reconstruction

License

Notifications You must be signed in to change notification settings

jfzhuang/IFR

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

IFR

This repository is the official implementation of "Semi-Supervised Video Semantic Segmentation with Inter-Frame Feature Reconstruction" (accepted by CVPR 2022). It is designed for semi-supervised video semantic segmentation task.

Install & Requirements

The code has been tested on pytorch=1.8.2 and python3.8. Please refer to requirements.txt for detailed information.

To Install python packages

pip install -r requirements.txt

Download Pretrained Weights

mkdir ./IFR/pretrained
cd ./IFR/pretrained
# download resnet18 imagenet pretrained weight
wget http://sceneparsing.csail.mit.edu/model/pretrained_resnet/resnet18-imagenet.pth
# download resnet101 imagenet pretrained weight
wget http://sceneparsing.csail.mit.edu/model/pretrained_resnet/resnet101-imagenet.pth

Data preparation

You need to download the Cityscapes datasets.

Your directory tree should be look like this:

./IFR/data
├── cityscapes
│   ├── gtFine
│   │   ├── train
│   │   └── val
│   └── leftImg8bit_sequence
│       ├── train
│       └── val

Prepare Downsample Dataset

Generated downsample dataset would be saved in ./data

cd ./IFR
python tools/data_downsample.py

Stage One Training of Accel

For example, train image segmentation model on 2 GPUs. Checkpoints would be saved in ./IFR/work_dirs.

# train PSP18 baseline model
cd ./IFR/exp/sup_30_res18/scripts
bash train.sh
# train PSP101 baseline model
cd ./IFR/exp/sup_30_res101/scripts
bash train.sh
# train PSP18 IFR model
cd ./IFR/exp/IFR_30_res18/scripts
bash train.sh
# train PSP101 IFR model
cd ./IFR/exp/IFR_30_res101/scripts
bash train.sh

Stage Two Training of Accel

For example, train Accel18 on 2 GPUs. Checkpoints would be saved in ./Accel/work_dirs.

mkdir ./Accel/work_dirs
# train Accel18 with baseline model
cd ./Accel/exp/accel18_30_sup/script
bash train.sh
# train Accel18 with IFR model
cd ./Accel/exp/accel18_30_IFR/script
bash train.sh

About

[CVPR'22] Semi-Supervised Video Semantic Segmentation with Inter-Frame Feature Reconstruction

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published