Skip to content

Installation.md

Latest commit

 

History

History
81 lines (57 loc) · 2.07 KB

Installation.md

File metadata and controls

81 lines (57 loc) · 2.07 KB
Raw

Installation

Requirements

  • Linux (tested on Ubuntu 24.04)
  • Python 3.7
  • Miniconda
  • PyTorch
  • CUDA 11.7

Create Anaconda Environment from yml

in the directory of BTS:

cd coperception
conda env create -f environment.yml
conda activate coperception

CUDA

conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia

Install CoPerception Library

This installs and links coperception library to code in ./coperception directory.

pip install -e .

Dataset Preparation

Please download and unzip the parsed dataset of CoSwarm.

Specifying Dataset

Link the test split of CoSwarm dataset in the default value of argument data

# BTS/coperception/tools/det/BTS/BTS_util.py
parser.add_argument("-d", "--data", default="/{your location}/dataset/CoSwarm-det/test", type=str, help="The path to the preprocessed sparse BEV training data", )

in the test folder data are structured like:

test
├──agent_0
├──agent_1
├──agent_2
├──agent_3
├──agent_4
├──agent_5
    ├──8_0
	    ├──0.npy		
    ...

Specifying Victim Detection Model Checkpoint

Link the checkpoint location in the default value of argument resume

Please download pre-trained weights and save them in BTS/coperception/tools/det/runs/resume/max/with_cross folder.

# BTS/coperception/tools/det/BTS/BTS_util.py
parser.add_argument("--resume", default="/{your location}/BTS/coperception/tools/det/runs/resume/max/with_cross/epoch_50.pth", type=str, help="The path to the saved model that is loaded to resume training", )

Specifying The Log Path

If you need log, don't forget to specify the log path in addition to --log.

# BTS/coperception/tools/det/BTS/BTS_util.py
parser.add_argument("--logpath", default="/{your location}/BTS/coperception/logs", help="The path to the output log file")