/Hybrid-EfficientNet-Swin-Transformer

An implementation of hybrid efficientnet swin transformer model.

Apache License 2.0Apache-2.0

Hybrid EfficientNet Swin Transformer (HENetSwinT)

Figure: Grad-CAMs. Left (Input), Middle (CNN), Right (CNN + Swin Transformer).

Code

This repo provides and unofficial implementation of Hybrid EfficientNet-Swin-Transformer in TensorFLow 2. Reference paper: https://arxiv.org/pdf/2110.03786.pdf.

  1. TF.Keras: Hybrid EfficientNet Swin Transformer TPU
  2. TF: Hybrid EfficientNet Swin-Transformer : GradCAM

DISCLAIMER

  • The implementation of swin-transformer in tf.keras is mostly borrowed from VcampSoldiers/Swin-Transformer-Tensorflow and keras-code-examples.
  • Please note, none of the un-official implementation is efficient enough; there can be some possible bugs. So, consider it if you use this. Please refer to the official implementation in pytorch in that case.