https://github.com/semihdervis/vit-image-classification-with-any-images
Training (fine-tuning) and evaluating Vision Transformer (ViT) models on any image dataset
https://github.com/semihdervis/vit-image-classification-with-any-images
computer-vision image-classification vision-transformer vision-transformer-image-classification vit
Last synced: 7 months ago
JSON representation
Training (fine-tuning) and evaluating Vision Transformer (ViT) models on any image dataset
- Host: GitHub
- URL: https://github.com/semihdervis/vit-image-classification-with-any-images
- Owner: semihdervis
- License: mit
- Created: 2024-08-25T22:07:39.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-08-30T15:00:40.000Z (about 1 year ago)
- Last Synced: 2025-01-30T12:27:09.716Z (9 months ago)
- Topics: computer-vision, image-classification, vision-transformer, vision-transformer-image-classification, vit
- Language: Python
- Homepage:
- Size: 27.3 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# ViT-Image-Classification-with-Any-Images
## Overview
This repository provides a pipeline for fine-tuning a Vision Transformer (ViT) model on custom image datasets using Hugging Face's Transformers library. The code is designed to offer flexibility in dataset management, model fine-tuning, and inference, making it easy to adapt the ViT model to various image classification tasks
## Setup
### Clone the Repository
```bash
git clone https://github.com/semihdervis/ViT-Image-Classification-with-Any-Images.git
cd ViT-Image-Classification-with-Any-Images
```
### Install Requirements
Ensure you have Python 3.8+ installed. Install the necessary packages using `pip`:
```bash
pip install -r requirements.txt
```
## Usage
### Training the Model
1. **Set Dataset and Output Directory:**
- Replace `DATASET_PATH` in [`train.py`](train.py) with the path to your image dataset.
- Set `OUTPUT_DIR` to your desired model output directory.
2. **Run Training:**
```bash
python train.py
```
### Testing the Model with a Single Image
1. **Set Model and Image Paths:**
- In [`test_model_with_single_image.py`](test_model_with_single_image.py), replace `MODEL_PATH` with the path to your trained model.
- Replace `IMAGE_PATH` with the path to the image you want to classify.
2. **Run the Inference Script:**
```bash
python test_model_with_single_image.py
```
### Testing the Model with Video Capture
1. **Set Model Path:**
- In [`test_model_with_video_capture.py`](test_model_with_video_capture.py), replace `MODEL_PATH` with the path to your trained model.
2. **Run the Video Capture Script:**
```bash
python test_model_with_video_capture.py
```
## License
This project is licensed under the [MIT License](LICENSE).