social_lstm_keras_tf

2019/01/22 - now Please See renew branch.

Social LSTM implementation with Keras (and TensorFlow as backend)
NOTE: experimental implementation

Requirements (show primary packages only)

• Python 3.6.0+
• Keras 2.1.5+
• TensorFlow 1.6.0+

Usage

1. Preparation

Set dataset attribute of the config files in configs/.

2. Training

Run train_social_model.py.

python train_social_model.py --config path/to/config.json [--out_root OUT_ROOT]

3. Testing

Run evaluate_social_model.py

python train_social_model.py --trained_model_config path/to/config.json --trained_model_file path/to/trained_model.h5

Restrictions

• work only on batch size = 1
• require much RAM (use almost all 16GB in my environment)

References

• Alexandre Alahi, Kratarth Goel, Vignesh Ramanathan, Alexandre Robicquet, Li Fei-Fei, Silvio Savarese. Social LSTM: Human Trajectory Prediction in Crowded Spaces. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 961-971
• My implementation based on: Social LSTM using TensorFlow