Implementation of double-prong ConvLSTM model for occupancy grid prediction as in "Double-Prong ConvLSTM for Spatiotemporal Occupancy Prediction in Dynamic Environments" (arXiv) by Maneekwan Toyungyernsub, Masha Itkina, Ransalu Senanayake, and Mykel J. Kochenderfer. Accepted to the IEEE The International Conference on Robotics and Automation (ICRA), Xi'an, 2021.
The LiDAR data used in the experiments are obtained from Waymo. We use their 31 training folders and split the data into training, validation, and test sets. The downloaded training folders should be put under the directory: DATA_waymo/. We also make use of the waymo-open-dataset github repository to interface with the downloaded data. Please put the downloaded waymo_open_dataset folder under the CODES_data_generation/ directory.
- python 3.6.9
- tensorflow-gpu (1.13.1)
- tensorboard (1.13.1)
- Keras (2.2.4)
- alt-model-checkpoint (2.0.2)
- numpy (1.18.1)
- hickle (3.4.5)
To generate the evidential grids, please run the following script from the CODES_data_generation/ directory:
generate_data.sh
Run the following script from CODES_prediction/double_prong_model/ directory to train the model:
train.sh
The input arguments can be modified in the train.sh script to change the number of gpus used for training and the output file name.