Dry Fruit Grade Classification

A deep learning model to classify images of dry fruits (such as almonds and cashews) into different quality grades (for example, Grade A and Grade B).

Key Features

• ResNet50-based deep learning model for high-accuracy image classification
• Two-stage transfer learning strategy (feature extraction & fine-tuning)
• 99.7% validation accuracy on augmented dataset
• Large dataset: 850 original images augmented to 42,600 images
• Interactive web deployment via Hugging Face Spaces
• Supports multiple dry fruit types (almonds, cashews, etc.) and quality grades (Grade A, Grade B)

Technologies Used

• Python
• TensorFlow
• Convolutional Neural Networks - ResNet50
• Streamlit for web app deployment
• Image processing techniques for data augmentation

Model Results

• Training Accuracy: ~99.7% validation accuracy achieved through two-stage training
  • Stage 1: Feature extraction with frozen ResNet50 backbone (~99.4% accuracy)
  • Stage 2: Fine-tuning entire model with low learning rate (1e-5)
• Inference Performance: Real-world testing shows confidence scores frequently above 75%
• Dataset: 42,600 augmented training images from 850 original high-resolution (720×720) images
• Data Augmentation: Rotations, brightness/contrast adjustments, and noise injection for robust model generalization

Deployment

The trained ResNet50 model is deployed as an interactive web application on Hugging Face Spaces, providing:

• Easy Access: Browse interface for uploading dry fruit images directly in your web browser
• Real-time Predictions: Instant classification with predicted quality grade (e.g., Grade A, Grade B)
• Confidence Scores: Model outputs confidence percentages for each prediction
• No Setup Required: Test the model without installing dependencies or running local code
• Live Demo: Access the interactive application