# CV-Image-Classification
**Repository Path**: xyfjason/cv-image-classification
## Basic Information
- **Project Name**: CV-Image-Classification
- **Description**: Reproduce image classification models with PyTorch.
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No
## Statistics
- **Stars**: 0
- **Forks**: 0
- **Created**: 2022-03-08
- **Last Updated**: 2023-02-15
## Categories & Tags
**Categories**: Uncategorized
**Tags**: None
## README
# Image-Backbones-Implementations
My implementations of image backbones with PyTorch.
## Training
```shell
python main.py [-n NAME] [-c FILE] [--opts KEY1 VALUE1 KEY2 VALUE2]
```
- To train on multiple GPUs, replace `python` with `torchrun --nproc_per_node NUM_GPUS`.
- An experiment directory will be created under `./runs/` for each run, which is named after `NAME`, or the current time if `NAME` is not specified. The directory contains logs, checkpoints, tensorboard, etc.
For example, to train resnet18 on CIFAR-10:
```shell
python main.py -c ./configs/resnet18_cifar10.yaml
```
## Results
### CIFAR-10
| models |
#params |
MACs |
acc@1(%) |
| VGG-11 |
9.20M |
0.15G |
90.97 |
| VGG-19 (BN) |
20.0M |
0.40G |
94.00 |
| ResNet-18 |
11.2M |
5.59G |
95.64 |
| PreActResNet-18 |
11.2M |
5.59G |
95.45 |
| ResNeXt-29 (32x4d) |
4.78M |
6.90G |
95.16 |
| SE-ResNet-18 |
11.3M |
5.59G |
95.65 |
| CBAM-ResNet-18 |
11.3M |
5.59G |
95.49 |
| MobileNet |
3.22M |
0.48G |
92.09 |
| ShuffleNet 1x (g=8) |
0.91M |
0.50G |
92.82 |
| ViT-Tiny/4 |
5.36M |
0.37G |
85.66 |
Note: MACs are calculated by [fvcore](https://github.com/facebookresearch/fvcore) library.
All the ConvNets are trained with the following settings:
- training duration: 64k steps
- batch size: 256
- learning rate: start with 0.1, end with 0.001 using a cosine annealing scheduler, no warm-up
- optimizer: SGD, weight decay 5e-4, momentum 0.9
The ViTs are trained with the following settings:
- training duration: 64k steps
- batch size: 512
- learning rate: start with 0.001, end with 0.00001 using a cosine annealing scheduler, no warm-up
- optimizer: Adam, weight decay 5e-5, betas (0.9, 0.999)