Skip to content

MediaBrain-SJTU/FreeAlign

Repository files navigation

FreeAlign (ICRA2024)

Robust Collaborative Perception without External Localization and Clock Devices

Original1

Paper

Installation

I recommend you visit Installation Guide to learn how to install this repo. Our code repository is based on the CoAlign Environment.

Or you can refer to OpenCOOD data introduction and OpenCOOD installation guide to prepare data and install the envirnment. The installation is totally the same as OpenCOOD.

The api of camera collaboration and the implementation Lift-Splat will be released in March. OPV2V, V2XSet and DAIR-V2X are supported. Pay attention to the repo if you need.

Note that I update the AP calculation (sorted by confidence) and add data augmentations (reinitialize) in this codebase, so the result will be higher than that reported in the current paper. I retrain all the models and would update the paper to the final version before March. Then I will remove this paragraph.

Complemented Annotations for DAIR-V2X-C

Originally DAIR-V2X only annotates 3D boxes within the range of camera's view in vehicle-side. We supplement the missing 3D box annotations to enable the 360 degree detection. With fully complemented vehicle-side labels, we regenerate the cooperative labels for users, which follow the original cooperative label format.

Download: Google Drive

Website: Website

Running Guidance

Download: Baidu Drive Down load the logs folder with your checkpoint and run inference with

python opencood/tools/inference.py --model_dir $CKPT_PATH

You can modify the config.yaml in the CKPT_PATH to see the performance gap between our method and previous methods. The key parameters include "gt_correct"(enable freealign) and noise_setting.

Acknowlege

This project is impossible without the code of CoAlign, OpenCOOD, g2opy and d3d!

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published