Environment setup

Clone the repo: git clone https://github.com/cmcin019/AV_Image_Analysis.git

1. Install conda for Linux

   wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
   bash Miniconda3-latest-Linux-x86_64.sh -b -p $HOME/miniconda
   $HOME/miniconda/bin/conda init bash
   

2. Create conda environment

   conda env create -f conda_env.yml
   conda activate AV_ENV
   
   cd AV_Image_Analysis
   

3. Create checkpoints folder

   mkdir checkpoints
   

4. Download checkpoints from add them to the checkpoints folder PSPNET https://drive.google.com/file/d/1ydBpkFAZ0CX7BD3iALGTr03UUUd_sfXn/view?usp=sharing

   DEPTH https://drive.google.com/file/d/13XS8X5p_mS_-SBuEsw1Sw7yGVUvYg0tY/view?usp=sharing

5. Download big lama folder and add it to the repository Big-LAMA https://drive.google.com/drive/folders/17l_wIZhZ-YWi_MUtCiz4164TnNr96Xxx?usp=sharing

6. Install mmcv and depth library

   pip install mmcv-full==1.3.13 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.8.0/index.html
   pip install -e .
   

7. Save path

   export TORCH_HOME=$(pwd) && export PYTHONPATH=.
   

Run

Compute SEGMENTATION and INPAINTING videos

python inpainting_test.py model.path=$(pwd)

Compute DEPTH video

python tools/test.py configs/depthformer/depthformer_swinl_22k_w7_kitti.py \
    checkpoints/depthformer_swinl_22k_kitti.pth \
    --show-dir road_depth

Compute road detection example

python road_detect_test.py

Acknowledgments

• Segmentation form MMSegmentation.
• Inpainting from LAMA
• Depth Estimation from Monocular-Depth-Estimation