Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/WenbinLee/DN4
Pytorch code of "Revisiting Local Descriptor based Image-to-Class Measure for Few-shot Learning", CVPR 2019.
https://github.com/WenbinLee/DN4
few-shot-learning image-to-class local-descriptors
Last synced: about 2 months ago
JSON representation
Pytorch code of "Revisiting Local Descriptor based Image-to-Class Measure for Few-shot Learning", CVPR 2019.
- Host: GitHub
- URL: https://github.com/WenbinLee/DN4
- Owner: WenbinLee
- License: other
- Created: 2019-03-25T21:24:54.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2023-06-20T02:12:30.000Z (over 1 year ago)
- Last Synced: 2024-07-31T23:45:10.312Z (4 months ago)
- Topics: few-shot-learning, image-to-class, local-descriptors
- Language: Python
- Homepage:
- Size: 5.58 MB
- Stars: 190
- Watchers: 10
- Forks: 43
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-few-shot-meta-learning - code - official (PyTorch)
README
# DN4 in PyTorch (2023 Version)
We provide a PyTorch implementation of DN4 for few-shot learning.
If you use this code, please cite:
[Revisiting Local Descriptor based Image-to-Class Measure for Few-shot Learning](http://cs.nju.edu.cn/rl/people/liwb/CVPR19.pdf).
[Wenbin Li](https://cs.nju.edu.cn/liwenbin/), Lei Wang, Jinglin Xu, Jing Huo, Yang Gao and Jiebo Luo. In CVPR 2019.
## Prerequisites
- Linux
- Python 3.8
- Pytorch 1.7.0
- GPU + CUDA CuDNN
- pillow, torchvision, scipy, numpy
## Getting Started
### Installation
- Clone this repo:
```bash
git clone https://github.com/WenbinLee/DN4.git
cd DN4
```
- Install [PyTorch](http://pytorch.org) 1.7.0 and other dependencies.
### Datasets
[Caltech-UCSD Birds-200-2011](https://data.caltech.edu/records/20098), [Standford Cars](https://ai.stanford.edu/~jkrause/cars/car_dataset.html), [Standford Dogs](http://vision.stanford.edu/aditya86/ImageNetDogs/main.html), [*mini*ImageNet](https://arxiv.org/abs/1606.04080v2) and [*tiered*ImageNet](https://arxiv.org/abs/1803.00676) are available at [Google Drive](https://drive.google.com/drive/u/1/folders/1SEoARH5rADckI-_gZSQRkLclrunL-yb0) and [百度网盘(提取码：yr1w)](https://pan.baidu.com/s/1M3jFo2OI5GTOpytxgtO1qA).
### miniImageNet Few-shot Classification
- Train a 5-way 1-shot model based on Conv64:
```bash
python Train_DN4.py --dataset_dir ./path/to/miniImageNet --data_name miniImageNet --encoder_model Conv64F_Local --way_num 5 --shot_num 1
```
- Train a 5-way 1-shot model based on ResNet12:
```bash
python Train_DN4.py --dataset_dir ./path/to/miniImageNet --data_name miniImageNet --encoder_model ResNet12 --way_num 5 --shot_num 1
```
- Test the model (specify the dataset_dir, encoder_model, and data_name first):
```bash
python Test_DN4.py --resume ./results/SGD_Cosine_Lr0.05_DN4_Conv64F_Local_Epoch_30_miniImageNet_84_84_5Way_1Shot/ --encoder_model Conv64F_Local
```
## Latest results on miniImageNet (2023)
(Compared to the originally reported results in the paper. * denotes that ResNet256F is used.)
Method
Backbone
5-way 1-shot
5-way 5-shot
2019 Version
2023 Version
2019 Version
2023 Version
DN4
Conv64F_Local
51.24
51.97
71.02
73.19
ResNet12
54.37*
61.23
74.44*
75.66
- The results on the miniImageNet dataset reported in the orinigal paper:
## Citation
If you use this code for your research, please cite our paper.
```
@inproceedings{DN4_CVPR_2019,
author = {Wenbin Li and
Lei Wang and
Jinglin Xu and
Jing Huo and
Yang Gao and
Jiebo Luo},
title = {Revisiting Local Descriptor Based Image-To-Class Measure for Few-Shot Learning},
booktitle = {{IEEE} Conference on Computer Vision and Pattern Recognition (CVPR)},
pages = {7260--7268},
year = {2019}
}
```