fruit application

Key technical highlights of the project: Developed with TensorFlow and MobileNet (a lightweight CNN architecture ideal for mobile devices, using transfer learning from pre-trained weights on ImageNet). Multi-class classification: fresh vs. rotten for each fruit type. Data sourced from public datasets (e.g., Kaggle fresh/rotten fruits collections) combined with additional images for better diversity and real-world applicability. Preprocessing included resizing images to fit MobileNet's input size, plus aggressive data augmentation (rotations, flips, etc.) to handle variations in lighting, angles, and backgrounds. The model was trained end-to-end and delivered an outstanding 98.5% accuracy on the test set. It earned a Very Good (B) grade — and it has clear practical value: helping minimize food waste in agriculture, supermarkets, or even at home by enabling fast, automated freshness checks.