# Mask-CDNet

**Repository Path**: lwq252713/Mask-CDNet

## Basic Information

- **Project Name**: Mask-CDNet
- **Description**: Change detection by pixel-level and semantic filtering using deep learning
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2020-11-18
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# mask-CDnet
Change detection by pixel-level and semantic filtering using deep learning. 

## Requirements
The program running on anaconda3 on Linux, which can be installed trrough https://www.anaconda.com/download/#linux

After installed the Anaconda3, please install following packages:
```
# pytorch
pip install pytorch

# OpenCV for python
pip install pyopencv

# python dense CRF
pip install pydensecrf

# Fast, Extensible Progress Meter
pip install tqdm
```


## Download pre-trained model
The pre-trained models used by this program can be downloaded at:


You can also execute the shell `download_pretrained_model.sh` to obtain it.


## Inference
```
CUDA_VISIBLE_DEVICE=0 python main.py \
    --inference --model FlowNet2S --save_flow \
    --inference_dataset AICD --inference_dataset_root ./AICD \
    --resume ./result/FlowNet2S_checkpoint.pth.tar \
    --inference_batch_size 2
```


## Training
```
CUDA_VISIBLE_DEVICE=1 python main.py \
    --training --model MaskCDnet \
    --loss MaskCDNet_loss --optimizer=Adam --optimizer_lr=1e-5 \
    --save_flow \
    --training_dataset AICD --training_dataset_root ./AICD --resume ./model/FlowNet2-S_checkpoint.pth.tar \
    --training_batch_size 2 --total_epochs 300
```