pose & bounding box prediction with intention

An open source code to predict human pose and bounding box as well as intention in JAAD dataset.

Contents

---

• Repository Structure
• Proposed Method
• Results
• Installation
• Dataset
• Training/Testing

Repository structure:

---

├── bbox-prediction-in-JAAD-with-intention         : bounding box & intention project repository
        ├── train.py                : Script for training including all necessary functions such as visualization and network  
        ├── test.py                 : Script for testing.  
        ├── DataLoader.py           : Script for data pre-processing and loader. 

├── pose-prediction-in-JAAD-with-intention         : pose & intention project repository
        ├── train.py                : Script for training including all necessary functions such as visualization and network  
        ├── test.py                 : Script for testing.  
        ├── DataLoader.py           : Script for data pre-processing and loader. 

Proposed method

---

We used a simple LSTM network similar to Pedestrian Intention Prediction: A Multi-task Perspective to predict both pose and bounding boxes as well as intention.

Results

---

In the below visualizations, we have shown observed frames in blue, ground truth future frames in green, and the predicted future frames in red. In addition, C and NC represent human crossing and non-crossing respectively.

blue: observation frames

red: predicted future frames

green: ground truth future frames

While our model can very accurately predict bounding boxes and intentions, it doesn't perform very well on the prediction of poses.

Installation:

---

Start by cloning this repositiory:

git clone https://github.com/vita-epfl/pose-intention.git
cd pose-intention

Create and activate virtual environment:

pip install --upgrade virtualenv
virtualenv -p python3 <venvname>  
source <venvname>/bin/activate  
pip install --upgrade pip

And install the dependencies:

pip install -r requirements.txt

Dataset:

We provided a preprocessed and balanced subset of JAAD dataset in each project's directory so that it is not needed to download and preprocess the original JAAD dataset.

Training/Testing:

Open train.py and test.py and change the parameters in the args class depending on the paths of your files. Start training the network by running the command:

python3 train.py

Test the trained network by running the command:

python3 test.py

Citation

@inproceedings{bouhsain2020pedestrian,
title={Pedestrian Intention Prediction: A Multi-task Perspective},
 author={Bouhsain, Smail and Saadatnejad, Saeed and Alahi, Alexandre},
  booktitle = {European Association for Research in Transportation  (hEART)},
  year={2020},
}