https://github.com/pprp/autorf
AutoRF: Learning Receptive Fields with Spatial Pooling
https://github.com/pprp/autorf
Last synced: 7 months ago
JSON representation
AutoRF: Learning Receptive Fields with Spatial Pooling
- Host: GitHub
- URL: https://github.com/pprp/autorf
- Owner: pprp
- License: mit
- Created: 2022-01-08T02:23:23.000Z (almost 4 years ago)
- Default Branch: master
- Last Pushed: 2023-05-18T01:35:44.000Z (over 2 years ago)
- Last Synced: 2025-01-31T13:43:55.390Z (9 months ago)
- Language: Python
- Homepage:
- Size: 330 KB
- Stars: 3
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# AutoRF: Auto Learning Receptive Fields with Spatial Pooling
Accepted by International Conference on MultiMedia Modeling 2023
## Introduction
This repository contains the source code of the paper: AutoRF: Auto Learning Receptive Fields with Spatial Pooling.
## Requirements
- Python 3.7+
- Cuda 11
- tensorboardX
- torch
- torchvision
- graphviz
- numpy
- thop
- timm
## Usage
### 1. Clone the repository
```shell
git clone https://github.com/pprp/AutoRF.git
```
### 2. Search
```shell
python search.py --data /path/to/data --dataset cifar10 --primitives fullpool --model_name rf_p5
```
### 3. Retrain
```shell
python tools/retrain.py --model_base 'rf_resnet20' \
--dataset 'cifar100' \
--cutout \
--cutout_length 8 \
--arch P1 \
--batch_size 128 \
--label_smooth \
--epochs 200 \
--scheduler 'steplr' \
--model_name 'rf_resnet20_cifar100_cutout8_bs128_ls_steplr'
```
## Citation
```
@inproceedings{Dong2023AutoRFAL,
title={AutoRF: Auto Learning Receptive Fields with Spatial Pooling},
author={Peijie Dong and Xin Niu and Zimian Wei and Hengyue Pan and Dongsheng Li and Zhen Huang},
booktitle={Conference on Multimedia Modeling},
year={2023}
}
```