# 🧠 Image Classification AI Model (CIFAR-100)

This repository contains a Vision Transformer (ViT)-based AI model fine-tuned for **image classification** on the CIFAR-100 dataset. The model is built using `google/vit-base-patch16-224`, quantized to **FP16** for efficient inference, and delivers high accuracy in multi-class image classification tasks.

---

## 🚀 Features

- 🖼️ **Task**: Image Classification  
- 🧠 **Base Model**: `google/vit-base-patch16-224` (Vision Transformer)  
- 🧪 **Quantized**: FP16 for faster and memory-efficient inference  
- 🎯 **Dataset**: 100 fine-grained object categories  
- ⚡ **CUDA Enabled**: Optimized for GPU acceleration  
- 📈 **High Accuracy**: Fine-tuned and evaluated on validation split  

---

## 📊 Dataset Used

**Hugging Face Dataset**: [`tanganke/cifar100`](https://huggingface.co/datasets/tanganke/cifar100)

- **Description**: CIFAR-100 is a dataset of 60,000 32×32 color images in 100 classes (600 images per class)
- **Split**: 50,000 training images and 10,000 test images
- **Categories**: Animals, Vehicles, Food, Household items, etc.
- **License**: MIT License (from source)

```python
from datasets import load_dataset

dataset = load_dataset("tanganke/cifar100")
```

## 🛠️ Model & Training Configuration
 
 - Model: google/vit-base-patch16-224

 - Image Size: 224x224 (resized from 32x32)

 - Framework: Hugging Face Transformers & Datasets

 - Training Environment: Kaggle Notebook with CUDA

 - Epochs: 5–10 (with early stopping)

 - Batch Size: 32

 - Optimizer: AdamW

 - Loss Function: CrossEntropyLoss

# ✅ Evaluation & Scoring

 - Accuracy: ~70–80% (varies by configuration)

 - Validation Tool: evaluate or sklearn.metrics

 - Metric: Accuracy, Top-1 and Top-5 scores

 - Inference Speed: Significantly faster after quantizationextractor = ViTFeatureExtractor.from_pretrained("google/vit-base-patch16-224")

# 🔍 Inference Example

```python
from PIL import Image
import torch

def predict(image_path):
    image = Image.open(image_path).convert("RGB")
    inputs = feature_extractor(images=image, return_tensors="pt").to("cuda")
    outputs = model(**inputs)
    logits = outputs.logits
    predicted_class = logits.argmax(-1).item()
    return dataset["train"].features["fine_label"].int2str(predicted_class)

print(predict("sample_image.jpg"))
```

# 📁 Folder Structure

📦image-classification-vit
 ┣ 📂vit-cifar100-fp16
 ┣ 📜train.py
 ┣ 📜inference.py
 ┣ 📜README.md
 ┗ 📜requirements.txt