https://github.com/jingew/critical-period-analysis

A computational explanation for face-learning behaviors by using deep artificial neural networks
https://github.com/jingew/critical-period-analysis

casia computational-biology dnn face-recognition knowledge-distillation machine-learning mtcnn python

Last synced: 3 months ago
JSON representation

A computational explanation for face-learning behaviors by using deep artificial neural networks

• Host: GitHub
• URL: https://github.com/jingew/critical-period-analysis
• Owner: JingeW
• License: mit
• Created: 2023-08-09T20:39:02.000Z (about 2 years ago)
• Default Branch: master
• Last Pushed: 2024-05-31T16:23:45.000Z (over 1 year ago)
• Last Synced: 2025-03-30T20:30:26.859Z (7 months ago)
• Topics: casia, computational-biology, dnn, face-recognition, knowledge-distillation, machine-learning, mtcnn, python
• Language: Python
• Homepage:
• Size: 5.26 MB
• Stars: 4
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Critical Period Analysis on Face Recognition

[![License: MIT](https://img.shields.io/badge/License-MIT-brightgreen.svg)](LICENSE)

We provide a full computational account for face-learning behaviors by using deep artificial neural networks and show that impaired face-learning can be restored when providing information within a critical period.

📃 [Read the Full Paper](https://www.cell.com/patterns/pdf/S2666-3899(23)00297-0.pdf)

## Requirements

1. numpy

2. pandas

3. scipy

4. pickle

5. matplotlib

6. sklearn

7. torchvision

## Data

The dataset used in this project is [CASIA-WebFace](https://www.kaggle.com/datasets/debarghamitraroy/casia-webface). \

You can download all the data used in this project with the [link](https://drive.google.com/file/d/1KxNCrXzln0lal3N4JiYl9cFOIhT78y1l/view).

Here are some examples of the data:

![Data Exp](Figs/Fig1-A.png)

## Disclaimer

This project is for academic research purposes only. The code in this repository is released under the MIT License.

If you use the data provided, please cite *Yi, Dong, et al. "Learning face representation from scratch." arXiv preprint arXiv:1411.7923 (2014).*

## Data Preprocessing

We applied foveate blurring to generate the data using different eye fixations.

![Data Prep](Figs/Fig1-C.png)

## Observations

We found the critical period played an important role in the learning phase of face recognition.

![The Grad-CAM for an example face](Figs/Fig2-E.png)

![The Grad-CAM group average across faces](Figs/Fig2-F.png)

## Recovery

We also proposed a method to recover the impairments caused during the Critical Period with an extremely low learning rate.

![KL-AT](Figs/Fig6-A.png)

With Knowledge Distillation and Attention Transfer, we recovered the accuracy of the impaired model.

 

The recovery is also confirmed by statistical metrics



## Citation

@article{wang2024critical,

  title={A critical period for developing face recognition},

  author={Wang, Jinge and Cao, Runnan and Chakravarthula, Puneeth N and Li, Xin and Wang, Shuo},

  journal={Patterns},

  volume={5},

  number={2},

  year={2024},

  publisher={Elsevier}

}