gorkaydemir / adapt Goto Github PK

Official implementation of the ICCV 2023 paper "ADAPT: Efficient Multi-Agent Trajectory Prediction with Adaptation"

License: MIT License

Python 44.40% Jupyter Notebook 55.60%

adapt's Introduction

ADAPT: Efficient Multi-Agent Trajectory Prediction with Adaptation

Paper | Webpage

This is the official implementation of the paper ADAPT: Efficient Multi-Agent Trajectory Prediction with Adaptation published in ICCV 2023.

Introduction

1. Clone this repository:

clone https://github.com/gorkaydemir/ADAPT.git
cd ADAPT

2. Create a conda environment and install required packages:

conda create -n adapt python=3.8
conda activate adapt
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch

3. Apply preprocessing to data and extract them into /path/to/data, following the process explained in dataset/README.md

Train

python run.py \
--ex_file_path /path/to/data/extended_ex_list \
--val_ex_file_path /path/to/data/eval.ex_list \
--model_save_path checkpoints/exp0 \
--static_agent_drop --scaling

Validation

python run.py --validate \
--ex_file_path /path/to/data/extended_ex_list \
--val_ex_file_path /path/to/data/eval.ex_list \
--model_save_path checkpoints/exp0 \
--checkpoint_path /path/to/checkpoint --use_checkpoint

You can download the pretrained model here.

How to Cite

@InProceedings{Aydemir2023ICCV,
        author = {Aydemir, G\"orkay and Akan, Adil Kaan and G\"uney, Fatma},
        title = {{ADAPT}: Efficient Multi-Agent Trajectory Prediction with Adaptation},
        booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},
        year      = {2023}}

adapt's People

Contributors

Stargazers

Watchers

Forkers

plusleizhou xtcl michaelh717 gaosz0755 sansangela dtpmh

adapt's Issues

Model train

Hello, I encountered the following problems when I used the pre-processed data for model training. How can I solve them?

$ python run.py --ex_file_path </home/miao/data/extended_ex_list> --val_ex_file_path </home/miao/data/eval.ex_list> --model_save_path </home/miao/checkpoints/exp0> --static_agent_drop --scaling
Fatal Python error: init_sys_streams: is a directory, cannot continue
Python runtime state: core initialized

Current thread 0x00007fe26418a180 (most recent call first):

isualization code

Hello, congratulations on achieving such great results, and thank you for sharing. Could you please open-source your visualization code?

Hardware requirements

First of all, congratulations on your paper being accepted by ICCV2023. Your paper is really helpful for my study, but due to hardware limitations, I only have two RTX 3090 graphics cards and 64GB of memory. Is this hardware configuration sufficient? Can you publish the hardware conditions of your experiment and the approximate time required for data preprocessing and model training?

cpu and gpu

Hello,

First of all, thank you for your outstanding contributions. During the training process, I monitored the usage of the CPU and GPU. While the CPU's 64GB of memory was being utilized, the GPU was only using a small amount, around 6GB. Have you encountered this issue before?

Visualization code/tutorials (gif)

Can you please upload visualization code what is present in your website .

INTERACTION dataset code training and validation

Hi,
First of all, your work is truly outstanding and has been incredibly helpful for our studies and research. We are also interested in replicating your work on other datasets to enhance its generalization capabilities. If possible, could you please provide the training , validation and visualizing code for the INTERACTION dataset?

Additionally, we are curious if the current model is suitable for adding predictions of speed and heading angle?

Best~

Process finished with exit code 136 (interrupted by signal 8:SIGFPE)
have you meet this problem?