🧠 Beans-Image-Classification-AI-Model

A fine-tuned image classification model trained on the Beans dataset with 3 classes: angular_leaf_spot, bean_rust, and healthy. This model is built using Hugging Face Transformers and the ViT (Vision Transformer) architecture and is suitable for educational use, plant disease classification tasks, and image classification experiments.

---

✨ Model Highlights

• 📌 Base Model: google/vit-base-patch16-224-in21k
• 📚 Fine-tuned: Beans dataset
• 🌿 Classes: angular_leaf_spot, bean_rust, healthy
• 🔧 Framework: Hugging Face Transformers + PyTorch
• 📦 Preprocessing: AutoImageProcessor from Transformers

---

🧠 Intended Uses

• ✅ Educational tools for training and evaluation in agriculture and plant disease detection
• ✅ Benchmarking vision transformer models on small datasets
• ✅ Demonstration of fine-tuning workflows with Hugging Face

---

🚫 Limitations

• ❌ Not suitable for real-world diagnosis in agriculture without further domain validation
• ❌ Not robust to significant background noise or occlusion in images
• ❌ Trained on small dataset, may not generalize beyond bean leaf diseases

---

📏 Input & Output

• Input: RGB image of a bean leaf (expected size 224x224)
• Output: Predicted class label — angular_leaf_spot, bean_rust, or healthy

---

🏋️‍♂️ Training Details

Attribute	Value
Base Model	`google/vit-base-patch16-224-in21k
Dataset	Beans Dataset (train/val/test)
Task Type	Image Classification
Image Size	224 × 224
Epochs	3
Batch Size	16
Optimizer	AdamW
Loss Function	CrossEntropyLoss
Framework	PyTorch + Transformers
Hardware	CUDA-enabled GPU

---

📊 Evaluation Metrics

Metric	Score
Accuracy	0.98
F1-Score	0.99
Precision	0.98
Recall	0.99

---

---

🚀 Usage

from transformers import AutoImageProcessor, AutoModelForImageClassification
from PIL import Image
import torch

model_name = "AventIQ-AI/Beans-Image-Classification-AI-Model"

processor = AutoImageProcessor.from_pretrained(model_name)
model = AutoModelForImageClassification.from_pretrained(model_name)
model.eval()

def predict(image_path):
    image = Image.open(image_path).convert("RGB")
    inputs = processor(images=image, return_tensors="pt").to(model.device)
    with torch.no_grad():
        outputs = model(**inputs)
    preds = torch.argmax(outputs.logits, dim=1)
    return model.config.id2label[preds.item()]

# Example
print(predict("example_leaf.jpg"))

---

• 🧩 Quantization
• Post-training static quantization applied using PyTorch to reduce model size and accelerate inference on edge devices.

---

🗂 Repository Structure

.
beans-vit-finetuned/
├── config.json               ✅ Model architecture & config
├── pytorch_model.bin         ✅ Model weights
├── preprocessor_config.json  ✅ Image processor config
├── special_tokens_map.json   ✅ (Auto-generated, not critical for ViT)
├── training_args.bin         ✅ Training metadata
├── README.md                 ✅ Model card

---

🤝 Contributing

Open to improvements and feedback! Feel free to submit a pull request or open an issue if you find any bugs or want to enhance the model.