Map-Aware Human Pose Prediction for Robot Follow-Ahead
To run the minimum code, clone the repository and go into the repository directory:
cd iros2024_poseForecasting
Install the requirements by running:
pip install -r requirements.txt
Then, download pre-processed dataset for GTA-IM and Real-IM and put them under folders in data/ directory (please follow the instructions in the Dataset section below). Next, download the model weights real_Path_GRU and gta_Path_GRU under results/real_Path_GRU/models and results/gta_Path_GRU/models directories, respectively.
You can run evaluation code on Real-IM dataset by:
python eval.py --cfg real_Path_GRU --epoch 3600
Similarly, to run on GTA-IM dataset:
python eval.py --cfg gta_Path_GRU --epoch 1200
To train the network, you can run the training script.
python train.py --cfg gta_Path_GRU --epoch 1200 --train_poseet --train_pathnet
To get permission to use datasets in the links, you will need to comply with a procedure.
Please follow the link to their repository and navigate under 'Requesting Dataset' and request their dataset. Then, after getting a response from them, please forward it to us so that we can provide you a subset of pre-processed dataset for testing and demonstration.
We provide a subset of the pre-processed version of the dataset in this link. We will publish the full version our dataset in the future.