This repository corresponds to the research documented in: RetinaNet Object Detector based on Analog-to-Spiking Neural Network Conversion.
The code stems from the library SNN_Toolbox, modifying it to address object detection problems. As a trade-off many side features of the original library have been removed. The conversion algorithm only accepts a Keras input model, and sticks to a 'temporal_mean_rate' encoding for the spiking activations and a 'constant_input_currents' encoding for the input images.
Configuration used:
- Python 3.8.6
- TensorFlow 2.3.1
- Keras 2.4.3