Skip to content

Project for Computer Vision course at Sapienza University of Rome

License

Notifications You must be signed in to change notification settings

Nhix00/EgoGesture_Recognition

Repository files navigation

Benchmark of Light Weight Models for EgoGesture Recognition

Static Badge Static Badge

Introduction

The proposed work primarily concentrates on offline gesture recognition as detailed in the paper. Our objective is to benchmark several lightweight models, such as ShuffleNet, and compare their performance to the heavier ResNeXt model featured in the study.

Dataset

In this work we used Egogesture dataset that is a multi-modal large scale dataset for egocentric hand gesture recognition. This dataset provides the test-bed not only for gesture classification in segmented data but also for gesture detection in continuous data.

Models

  • MobileNet: Optimized for mobile and embedded devices with depth-wise separable convolutions for high-performance image recognition
  • ShuffleNet: Minimizes computational complexity with channel shuffling, ideal for mobile and resource-constrained applications

We used pretrainet models on Jester dataset from this paper: link, they published their work on GitHub (link)

Results

The results of our work are shown in the next graphic.

Benchmark of the tested models

Contacts

If you have any questions, suggestions, or feedback, we'd love to hear from you! Here's how you can reach out:

About

Project for Computer Vision course at Sapienza University of Rome

Topics

Resources

License

Stars

Watchers

Forks