https://github.com/yinboc/prototypical-network-pytorch

A re-implementation of "Prototypical Networks for Few-shot Learning"
https://github.com/yinboc/prototypical-network-pytorch

few-shot-learning machine-learning pytorch

Last synced: 11 days ago
JSON representation

A re-implementation of "Prototypical Networks for Few-shot Learning"

• Host: GitHub
• URL: https://github.com/yinboc/prototypical-network-pytorch
• Owner: yinboc
• License: mit
• Created: 2018-07-07T04:51:23.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2020-03-26T07:13:45.000Z (over 4 years ago)
• Last Synced: 2024-04-10T06:06:50.325Z (3 months ago)
• Topics: few-shot-learning, machine-learning, pytorch
• Language: Python
• Homepage:
• Size: 174 KB
• Stars: 295
• Watchers: 6
• Forks: 61
• Open Issues: 3
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Lists

• awesome-few-shot-meta-learning - code (PyTorch)

README

        
# Prototypical Network

A re-implementation of [Prototypical Network](https://arxiv.org/abs/1703.05175).

With ConvNet-4 backbone on miniImageNet.

***For deep backbones (ResNet), see [Meta-Baseline](https://github.com/cyvius96/few-shot-meta-baseline).***

### Results

1-shot: 49.1% (49.4% in the paper)

5-shot: 66.9% (68.2% in the paper)

## Environment

* python 3

* pytorch 0.4.0

## Instructions

1. Download the images: https://drive.google.com/open?id=0B3Irx3uQNoBMQ1FlNXJsZUdYWEE

2. Make a folder `materials/images` and put those images into it.

`--gpu` to specify device for program.

### 1-shot Train

`python train.py`

### 1-shot Test

`python test.py` 

### 5-shot Train

`python train.py --shot 5 --train-way 20 --save-path ./save/proto-5`

### 5-shot Test

`python test.py --load ./save/proto-5/max-acc.pth --shot 5`