Skip to content

Latest commit

 

History

History
47 lines (42 loc) · 1.72 KB

prepare.md

File metadata and controls

47 lines (42 loc) · 1.72 KB

checkpoint

  • Download the panFPN model from google drive: panFPN_checkpoint and save to checkpoints/panFPN.pth
  • Download the vo model from google drive:vo_checkpoint and save to checkpoints/vkitti2_dy_train_semiv4_080000.pth

droidenv env

  1. Creating a new anaconda environment using the provided .yaml file. Use VO_Module/environment_novis.yaml to if you do not want to use the visualization
conda env create -f VO_Module/environment.yml
pip install evo --upgrade --no-binary evo
pip install gdown
  1. Compile the extensions (takes about 10 minutes)
python setup.py install
  1. video panoptic segmentation requirements. The Video panoptic segmentation module is based on Detectron2, you can install Detectron2 following the instructions.
conda activate droidenv
conda install pytorch==1.9.0 torchvision cudatoolkit=11.1 -c pytorch -c nvidia

python -m pip install -e VPS_Module
pip install git+https://github.com/cocodataset/panopticapi.git

vkitti 15-deg-left dataset

Expected dataset structure for Virtual_KITTI2

Virtual_KITTI2/
  Scene01/
    15-deg-left/
    15-deg-right/
    ...
    clone/
  Scene02/
  Scene06/
  Scene18/
  Scene20/
  • generate annotation for training and evaluating
conda activate droidenv
sh tools/datasets/generate_vkitti_datasets.sh
python tools/datasets/generate_dynamic_masks.py