# CLIP ViT Base Patch32 Fine-Tuned on PatchCamelyon (PCAM)

## Overview
This repository contains a fine-tuned version of the [CLIP ViT Base Patch32](https://huggingface.co/tanganke/clip-vit-base-patch32_pcam) model on the [PatchCamelyon (PCAM)](https://huggingface.co/datasets/1aurent/PatchCamelyon) dataset. The model is optimized for histopathological image classification.

---
tags:
- vision
- clip
- fine-tuned
- PatchCamelyon
- medical-imaging
license: apache-2.0
library_name: transformers
model_type: clip_vision_model
datasets:
- 1aurent/PatchCamelyon
---

## Model Details
- **Base Model**: `openai/clip-vit-base-patch32`
- **Dataset**: `PatchCamelyon`
- **Fine-tuned for**: Medical image classification (tumor vs. non-tumor)
- **Evaluation Results**:
  - **Train Accuracy**: 94.35%
  - **Validation Accuracy**: 95.16%
- **Hardware**: Trained on GPU-A100

## Training Performance
- **Train Loss**: 0.1520
- **Train Accuracy**: 94.35%
- **Validation Accuracy**: 95.16%

## Usage
### Installation
Ensure you have `transformers`, `torch`, and `safetensors` installed:
```bash
pip install transformers torch safetensors
```

### Loading the Model
```python
from transformers import CLIPProcessor, CLIPModel
import torch

model_path = "lens-ai/clip-vit-base-patch32_pcam_finetuned"
model = CLIPModel.from_pretrained(model_path)
processor = CLIPProcessor.from_pretrained(model_path)
```

### Running Inference
```python
from PIL import Image

image = Image.open("sample_image.png")
inputs = processor(images=image, return_tensors="pt")
outputs = model.get_image_features(**inputs)
```

## Evaluation
We plan to release additional metrics, including robustness evaluation with adversarial attacks in future updates.

## License
This model is released under the MIT License.

## Contact
For any questions, please reach out to **Venkata Tej** at [LensAI](https://huggingface.co/lens-ai).