094-parameter-efficient-long-tailed-recognition

https://github.com/szu-advtech-2023/094-parameter-efficient-long-tailed-recognition

Repository

https://github.com/SZU-AdvTech-2023/094-Parameter-Efficient-Long-Tailed-Recognition/blob/main/

# Reproduce experiments
These reproduce experiments are based on the Parameter-Efficient Long-Tailed Recognition http://arxiv.org/abs/2309.10019

## Overview of PEL

  


## Requirements

* Python 3.8
* PyTorch 2.0
* Torchvision 0.15
* Tensorboard

- Other dependencies are listed in [requirements.txt](requirements.txt).

To install requirements, run:

```sh
conda create -n pel python=3.8 -y
conda activate pel
conda install pytorch==2.0.0 torchvision==0.15.0 pytorch-cuda=11.7 -c pytorch -c nvidia
conda install tensorboard
pip install -r requirements.txt
```

We encourage installing the latest dependencies. If there are any incompatibilities, please change the dependencies with the specified version [requirements-with-version.txt](requirements-with-version.txt).

## Hardware

All experiments are conducted on the single GPU RTX3090.


## Quick Start on the CIFAR-100-LT dataset

```bash
# run PEL on CIFAR-100-LT (with imbalanced ratio=100)
python main.py -d cifar100_ir100 -m clip_vit_b16_peft 
```

By running the above command, you can automatically download the CIFAR-100 dataset and run the original PEL.

## Exploration experiments based on PEL

### LN+AdaptFormer

We keep layerNorm learnable when using Adaptformer fine-tuning pre-trained ViT.
```bash
# run LN+AdaptFormer on CIFAR-100-LT (with imbalanced ratio=100)
python main.py -d cifar100_ir100 -m clip_vit_b16_peft ln_tuning True
```

### VPT-S+Partial
Use vpt-shallow to improve partial AdaptFormer
```bash
# run VPT-S+Partial on CIFAR-100-LT (with imbalanced ratio=100)
python main.py -d cifar100_ir100 -m clip_vit_b16_peft vpt_shallow True partial 6
# run only partial
python main.py -d cifar100_ir100 -m clip_vit_b16_peft partial 6

```


## Other Dataset

Download the dataset [Places](http://places2.csail.mit.edu/download.html), [ImageNet](http://image-net.org/index), and [iNaturalist 2018](https://github.com/visipedia/inat_comp/tree/master/2018).

Put files in the following locations and change the path in the data configure files in [configs/data](configs/data):

- Places

```
Path/To/Dataset
 train
   airfield
|  |   00000001.jpg
|  |   ......
   ......
 val
    airfield
   |   Places365_val_00000435.jpg
   |   ......
    ......
```

- ImageNet

```
Path/To/Dataset
 train
   n01440764
|  |   n01440764_18.JPEG
|  |   ......
   ......
 val
    n01440764
   |   ILSVRC2012_val_00000293.JPEG
   |   ......
    ......
```

- iNaturalist 2018

```
Path/To/Dataset
 train_val2018
    Actinopterygii
   |   2229
   |  |   2c5596da5091695e44b5604c2a53c477.jpg
   |  |   ......
   |   ......
    ......
```

Owner

Citation (citation.txt)

@article{REPO094,
    author = "Shi, Jiang-Xin and Wei, Tong and Zhou, Zhi and Han, Xin-Yan and Shao, Jie-Jing and Li, Yu-Feng",
    journal = "arXiv preprint arXiv:2309.10019",
    title = "{Parameter-Efficient Long-Tailed Recognition}",
    year = "2023"
}

GitHub Events