[23.04.24] Update!
🤖 My Implementation of the approach described in below paper. Hope you got helped!
He, Tong, et al. "Bag of tricks for image classification with convolutional neural networks." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019.
https://arxiv.org/abs/1812.01187
Experiments for below methods introduced in this paper.
- Large Batch Training
- Cosine Learning Rate Decay
- Label Smoothing
- Transfer Learning
You can experiment upper methods using this repository.
GPU
NVIDIA RTX-3060-12GB, A6000-48GB, Colab-K80
CUDA
11.2
pytorch
1.9.1+cu111
torchvision
0.10.1+cu111
python
3.9.1
OS
Ubuntu 20.04, Ubuntu 18.04
Datasets
CUB200-2011, ImageNet-1K
(As I don't use special libraries, errors may not occur in other environments.)
Plese download the dataset from https://www.kaggle.com/datasets/coolerextreme/cub-200-2011. And split your data into train, test folders.
${YOUR_ROOT}/
|-- workspace
| |-- Bag-of-Tricks-for-Image-Classification # current repository.
|-- dataset
| |-- archive
| | |-- CUB_200_2011
| | | |-- CUB_200_2011
| | | | |-- train
| | | | |-- test
| | | | |-- ...
Please visit each directory's readme.md
!
This project is licensed under the terms of the MIT license.