Fresh vs Rotten Fruit Classification using Computer Vision
This project is based on the Kaggle dataset 'Fresh vs Rotten Fruit Images', which contains labeled images of fresh and rotten fruits such as apples, bananas, and oranges. The dataset is intended for image classification tasks where machine learning models are trained to distinguish between fresh and spoiled produce using visual cues.
Mentor Details:
Orna Bregman Amitai
Mentor Details:
Requirments:
- Build an image classification model to distinguish between fresh and rotten fruits.
- Explore data preprocessing and augmentation techniques to improve model robustness.
- Train and evaluate convolutional neural networks on the provided dataset.
- Analyze model performance and common misclassification patterns.
Problem Statement
Manual inspection of fruit quality is subjective, labor-intensive, and inefficient at scale. In agriculture, food supply chains, and retail, there is a growing need for automated systems that can reliably detect fruit freshness to reduce food waste, ensure quality control, and improve operational efficiency.
Project Objectives
- Build an image classification model to distinguish between fresh and rotten fruits.
- Explore data preprocessing and augmentation techniques to improve model robustness.
- Train and evaluate convolutional neural networks on the provided dataset.
- Analyze model performance and common misclassification patterns.
Technical Scope
- Image loading and preprocessing (resizing, normalization)
- Data augmentation techniques for image classification
- Training convolutional neural networks (CNNs)
- Model evaluation using accuracy, precision, recall, and confusion matrices
- Optional extensions: transfer learning and deployment as a simple inference app
Required Knowledge and Prerequisites
Core Requirements:
- Python programming
- Basic machine learning concepts
- Familiarity with deep learning frameworks such as TensorFlow or PyTorch
Recommended Background:
- Computer vision fundamentals
- Experience with CNN architectures
- Basic understanding of image datasets and data augmentation
Project Difficulty and Expected Level
Overall Difficulty: Beginners
This project is well-suited for:
Teams of 1–3 students
Expected Outcomes
- A trained image classification model capable of identifying fresh vs rotten fruits
- Quantitative evaluation results on a validation or test set
- A reproducible training and evaluation pipeline
- Practical experience applying computer vision to real-world quality control problems