TinyML-Based Ripeness Classification System for Mango and Banana Deployed on Raspberry Pi

View at publisher Summarise with VIRAT
Authors: Revathi GP, Sushanth S Naik, Vilas, Vidyasagar, Prasad B H
Cited by: 0
References: 8

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Accurately identifying the ripeness of fruits is important in agriculture and food supply chains. This paper presents a smart fruit ripeness classification system for mangoes and bananas using deep learning and TinyML. Two lightweight convolutional neural network (CNN) models based on MobileNetV2 are trained: a three-class mango classifier (Raw Mango, Ripe Mango, not a Mango) and a two-class banana classifier (Raw Banana, Ripe Banana). The models are trained on a balanced dataset of 3000 images with real world variation and improved using data augmentation and partial fine-tuning. The trained models are converted into TensorFlow Lite (.tflite) format and deployed on a Raspberry Pi with a camera, enabling real time classification. The mango classifier first checks if the fruit is a mango; if not, the image is passed to the banana classifier. The system achieved high classification accuracies of 99.30 % for mangoes and 99.00 % for bananas, validated through multiple folds to ensure robustness. This work demonstrates an efficient and portable and scalable Edge AI solution for fruit ripeness detection using TinyML.

Fields of Study & Subjects

Smart Agriculture and AI Advanced Neural Network Applications IoT and Edge/Fog Computing

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1061–1066
Publication date
21 January 2026
Language
EN