Few Shot Learning

Implementation of a few shot learning algorithm using Tensorflow.js with a mobilenetv2 backbone and a Knn-Classifier on top of it.

This project is based on the official tutorial: https://www.tensorflow.org/js/tutorials/transfer/image_classification

Usage

1. To classify image from disk: Run classifier/index.html
2. To classify image from webcam: Run classifier_camera.html

Demo

1. To classify images from disk use this link
2. To classify image from webcam use this link

Data

Example data is included wit 3 datasets

• Animals
  • Cats: 10 images
  • Dogs: 10 images
  • Lions: 10 images
• Animals_inference
  • Cats: 1 image
  • Dogs: 1 image
  • Lions: 1 image
  • Fake lions: 2 images
• Blur_detection
  • Blured: 5 images
  • Focused: 5 items
• Colors
  • Blue: 10 images
  • Green: 10 images
  • Red: 10 images