Real-time object detection on Android using YOLO with TensorFlow, detecting 20 object classes from the Pascal VOC dataset.
android-yolo is an Android application that implements real-time object detection using the YOLO (You Only Look Once) neural network with TensorFlow. It allows Android devices to detect 20 different object classes from the Pascal VOC dataset directly on the device without requiring cloud processing. The project was the first implementation of YOLO for TensorFlow on Android.
Android developers and researchers interested in implementing real-time computer vision applications on mobile devices, particularly those working with object detection and TensorFlow mobile deployments.
Developers choose android-yolo because it provides a working implementation of YOLO for Android with TensorFlow, offering real-time object detection capabilities directly on device with minimal setup. It's particularly valuable as it was the first such implementation and includes both source code and a standalone APK for immediate testing.
Real-time object detection on Android using the YOLO network with TensorFlow
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
This was the first project to bring YOLO object detection to TensorFlow on Android, making it a historical reference for mobile computer vision.
Provides a downloadable APK for easy installation without needing to compile or set up Android Studio, allowing quick testing.
Enables object detection directly on Android devices at around 2 FPS, eliminating the need for cloud processing and reducing latency.
Compatible with Android Studio and comes with source code that can be run out of the box, facilitating development and customization.
Only outputs one predicted bounding box per detection, severely limiting its ability to detect multiple objects simultaneously as intended in YOLO.
Hardcoded for the 20 Pascal VOC classes, with no built-in support for custom models or additional object categories.
Processes images at only about two frames per second on capable devices, which may not be sufficient for smooth real-time applications.
TensorFlow on Android does not support GPU acceleration for this implementation, capping performance on devices with capable hardware.