This repository contains the implementation of SpineNet-49S using Keras. SpineNet is a scale-permuted backbone model designed for efficient recognition and localization tasks. The model is based on the paper SpineNet: Learning Scale-Permuted Backbone for Recognition and Localization by Mingxing Tan et al.
SpineNet introduces a novel architecture for feature pyramids by learning scale-permuted networks. The model demonstrates state-of-the-art performance on recognition and localization tasks while being computationally efficient.
The SpineNet-49S variant is a compact version of the model, suitable for resource-constrained environments.
For more details, please refer to the original paper.
- Model Implementation: The SpineNet-49S architecture has been implemented in Keras.
Make sure you have the following installed:
- Python 3.8 or later
- TensorFlow 2.4 or later
- Keras 2.4 or later