Skip to content

Latest commit

 

History

History
72 lines (48 loc) · 1.5 KB

Installation.md

File metadata and controls

72 lines (48 loc) · 1.5 KB
Raw

Installation

Requirements

  • Linux (tested on Ubuntu 18.04)
  • Python 3.7
  • Anaconda
  • PyTorch
  • CUDA 11.7

Create Anaconda Environment from yml

in the directory of ROBOSAC:

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 detection dataset of V2X-Sim 2.0.

Specifying Dataset

Link the test split of V2X-Sim dataset in the default value of argument "data"

/{Your_location}/V2X-Sim-det/test

in the test folder data are structured like:

test
├──agent_0
├──agent_1
├──agent_2
├──agent_3
├──agent_4
├──agent_5
      ├──19_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 ROBOSAC/coperception/ckpt/meanfusion folder.

epoch_49.pth is the original victim model without adversarial training.

epoch_advtrain_49.pth is the PGD-trained model.