TorgTech Android TensorFlow Lite Apps

GPU-accelerated Android NDK examples for running TensorFlow Lite computer-vision models on-device. The repository is useful for Android developers evaluating native TFLite GPU Delegate builds for driver-monitoring and ADAS-style prototypes.

• Mati_DSMS/: camera-based face detection and face/iris landmark experiments using bundled MediaPipe TensorFlow Lite models.
• Tarzan_ADAS/: dense depth, segmentation, object detection, and face mesh examples packaged as an Android NDK app.
• Sahay_ADAS/: additional ADAS app source and assets covering dense depth, detection, segmentation, and face mesh models.
• third_party/: TensorFlow Lite build scripts and checked-in support code used by the Android app projects.

• Download and install Android NDK.

$ mkdir ~/Android/
$ mv ~/Download/android-ndk-r20b-linux-x86_64.zip ~/Android
$ cd ~/Android
$ unzip android-ndk-r20b-linux-x86_64.zip

• Download and install bazel.

$ wget https://github.com/bazelbuild/bazel/releases/download/3.1.0/bazel-3.1.0-installer-linux-x86_64.sh
$ chmod 755 bazel-3.1.0-installer-linux-x86_64.sh
$ sudo ./bazel-3.1.0-installer-linux-x86_64.sh

• run the build script to build TensorFlow Library

$ mkdir ~/work
$ git clone https://github.com/Akhilesh-Gogikar/TorgTech.git
$ cd TorgTech/third_party/
$ ./build_libtflite_r2.4_android.sh

(Tensorflow configure will start after a while. Please enter according to your environment)

$ ls -l tensorflow/bazel-bin/tensorflow/lite/

$ ls -l tensorflow/bazel-bin/tensorflow/lite/delegates/gpu/

Mati_DSMS includes its model assets in the repository. For the ADAS app assets, run the downloader that is checked into the target app:

$ cd ~/work/TorgTech/Tarzan_ADAS/app/src/main/assets/model
$ bash download.sh

Sahay_ADAS/app/src/main/assets/model/download.sh is the matching downloader for the Sahay asset folder when porting that source into an Android app.

• Download and install Android Studio.
• Start Android Studio.

$ cd ${ANDROID_STUDIO_INSTALL_DIR}/android-studio/bin/
$ ./studio.sh

• Install NDK 20.0 by SDK Manager of Android Studio.
• Open one checked-in Android Studio project from the quick build targets table, such as ~/work/TorgTech/Mati_DSMS or ~/work/TorgTech/Tarzan_ADAS.
• Build and Run.