🧠 Image Classification AI Model (CIFAR-100)

🚀 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

• 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)

🛠️ 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

📁 Folder Structure