Adapt-FuseNet: Context-aware Multimodal Adaptive Fusion of Face and Gait Features using Attention Techniques for Human Identification

Repository for the implementation of the paper "Adapt-FuseNet: Context-aware Multimodal Fusion of Face and Gait Features using Attention Techniques for Human Identification" presented in IJCB 2023.

The paper proposes a new fusion architecture which in combination with attention techniques improves Human Identification by obtaining face and gait of the subject.

Paper Link: Coming Soon

Environment

The code is written in Python 3.10.

Installation

• Download and Install Anaconda for virtual environment creation.
  

• Clone the repository

• All the necessary packages used in the study has been included in requirements.txt file and can be installed by running
  pip install -r requirements.txt in the repository folder

Dataset

The study employed the use of both CASIA-A and CASIA-B dataset for training the model.
Data Folder structure and local machine set up will be revealed in the future.

What's New?

• Sep 2023: Our paper won the IAPR Best Biometrics Student Paper Award 2023 in IJCB 2023
• Apr 2024: The paper is published as a conference paper in IEEE. Link is present in the Link to Paper section.

  Citation

  Pre-print

  Prakash, A., Nambiar, A., & Bernardino, A. (2023). Multimodal Adaptive Fusion of Face and Gait Features using Keyless attention based Deep Neural Networks for Human Identification. arXiv preprint arXiv:2303.13814.

  Conference Paper

  Prakash, A., Thejaswin, S., Nambiar, A., & Bernardino, A. (2023, September). Adapt-FuseNet: Context-aware Multimodal Adaptive Fusion of Face and Gait Features using Attention Techniques for Human Identification. In 2023 IEEE International Joint Conference on Biometrics (IJCB) (pp. 1-10). IEEE.

  Link to Paper

• Pre-print
  Multimodal Adaptive Fusion of Face and Gait Features using Keyless attention based Deep Neural Networks for Human Identification
  - Ashwin Prakash, Thejaswin S, Athira Nambiar, Alexandre Bernardino
• Conference Paper
  Adapt-FuseNet: Context-aware Multimodal Adaptive Fusion of Face and Gait Features using Attention Techniques for Human Identification
  - Ashwin Prakash, Thejaswin S, Athira Nambiar, Alexandre Bernardino