Skip to content

Latest commit

 

History

History
66 lines (47 loc) · 3.18 KB

get_started.md

File metadata and controls

66 lines (47 loc) · 3.18 KB

Get Started for V2X-Seq Dataset and Benchmark

SPD Dataset Usage

SPD is the successor to the DAIR-V2X-C dataset. While maintaining the overall structure, we have cleaned the dataset, annotated tracking IDs for each object, and provided vector maps.

SPD Benchmarks

We offer early fusion, late fusion, and middle fusion benchmarks like FF-Tracking for the VIC3D Tracking task. To learn about training and evaluating these benchmarks, visit the following link:

TFD Dataset Usage

The TFD dataset comprises trajectories, vector maps, and traffic light signals.

TFD Benchmarks

We provide various benchmarks, including PP-VIC, for solving Online-VIC Forecasting and Offline-VIC Forecasting tasks. Find basic guidance in the TFD Benchmark README. Detailed training and evaluation of Baselines with HiVT and TNT are as follows:

  • For training and evaluation of Baselines with HiVT, refer to the HiVT README.
  • For training and evaluation of Baselines with TNT, refer to the TNT README.

Example of PP-VIC Evaluation with HiVT

Here's how to evaluate PP-VIC with HiVT for solving the Online-VIC Forecasting task using the TFD-Example dataset.

  1. Create a conda environment and install dependencies as specified in HiVT:

    conda create -n HiVT python=3.8
    conda activate HiVT
    conda install pytorch==1.8.0 cudatoolkit=11.1 -c pytorch -c conda-forge
    conda install pytorch-geometric==1.7.2 -c rusty1s -c conda-forge
    conda install pytorch-lightning==1.5.2 -c conda-forge
  2. Install the Argoverse 1 API.

  3. Dataset Preparation:

    # Download TFD-Example into ./dataset/v2x-seq-tfd/V2X-Seq-TFD-Example
    bash tools/dataset_example_download.sh
    
    # export DATA_ROOT, change the DATA_ROOT to TFD-Example data root
    export DATA_ROOT=${PWD}'/dataset/v2x-seq-tfd/V2X-Seq-TFD-Example'
    
    # Merge Multiple Maps into One Map
    python tools/data_converter/maps_merge.py
    
    # Preprocess Cooperative-view Trajectories
    python tools/trajectory_fusion/fusion_for_prediction.py
    python tools/data_converter/tfd_argoverse_converter.py
  4. Evaluation:

    # parsers: GPU_ID, DATA_ROOT, CKPT
    # DATA_ROOT=../../dataset/v2x-seq-tfd/V2X-Seq-TFD-Example
    cd projects/HiVT_plugin
    bash tools/hivt_eval.sh 0 ${DATA_ROOT}/cooperative-vehicle-infrastructure/fusion_for_prediction ./checkpoints/online.ckpt