ArcFace_MATLAB

This is implementatoin of ArcFace(https://arxiv.org/abs/1801.07698) in MATLAB. In this demo, train and test pet images by using ArcFace. This network can

1. calculate the similarity between your pet images and each breed of dogs and cats.
2. calculate the similarity between your pet images and the other pet images.

Products used in development

• MATLAB R2021b
• Image Processing Toolbox
• Deep Learning Toolbox
• add on: Deep Learning Toolbox Model for ResNet-50 Network
• Parallel Computing Toolbox (if you want to use GPU in training)

Setup

dataset: Datasets is automatically downloaded from Oxford-IIIT Pet Datset(https://www.robots.ox.ac.uk/~vgg/data/pets/).

To Run:
・Train and Test:(GPU is recommanded)
　1．Open arcface_train.mlx
　2. Change directory to "arcface_demo"
　3. Execute Run
・Test:
　1．Open arcface_train.mlx
　2. Change directory to "arcface_demo"
　3. Go to "Test Model" Section
　　 If you want to use train model by run, remove commnet of "load dlnet_end.mat"
　4．Run section of "Test Model"

Getting Started

Please open arcface_train.mlx　and read instructions

Examples

To learn how to use this in testing workflows, see arface_train.mlx.

License

The license for Implementation-ArcFace-in-MATLAB is available in the license.txt file in this GitHub repository.

Include any other License information here, including third-party content using separate license agreements

Community Support

MATLAB Central

Copyright (c) 2022 Nobuaki Fukuchi.