Leveraging TensorFlow Lite and Camera2 API for Efficient Real-Time Object Detection in Android Apps Using Kotlin
- Version
- Download 3
- File Size 574.84 KB
- File Count 1
- Create Date 5 June 2025
- Last Updated 5 June 2025
Leveraging TensorFlow Lite and Camera2 API for Efficient Real-Time Object Detection in Android Apps Using Kotlin
Authors:
Aniruddha Arun Kharve, Monika Shinde
1Aniruddha Kharve Master of Computer Application & Trinity Academy of Engineering, Pune
2 Monika Shinde Master of Computer Application & Trinity Academy of Engineering, Pune
Abstract - This study uses Kotlin, Camera2 API, and TensorFlow Lite to design and construct an Android application for real-time object identification. The project aims to provide an efficient mobile solution that identifies and classifies objects through the phones camera in real time. To enhance accuracy and performance across various devices, the app integrates four pre-trained lightweight machine learning models: MobileNetV1, EfficientNet-Lite, EfficientNet-Lite1, and EfficientNet-Lite2. Additional features include image selection from the gallery, threshold-based object detection, and classification into categories like food, fashion, and electronics. The Android 14-compatible application implements modern runtime permission handling and supports smooth on-device ML operations. The results demonstrate satisfactory object recognition accuracy and usability, making it suitable for educational, lifestyle, and accessibility applications.
Key Words: Real-time Object Detection, TensorFlow Lite, Kotlin, Camera2 API, Android 14, MobileNet, EfficientNet.