This app is for indoor segmentation - indoor semantic segmentation for Android devices
using Deeplabv3 MobileNetV2 trained with ADE20K dataset.
The demo app is based on Tensorflow android sample app with bazel build.
Inference is done using the TensorFlow Android Inference
Interface, which may be built separately
if you want a standalone library to drop into your existing application. Object
tracking and efficient YUV -> RGB conversion are handled by
libtensorflow_demo.so.
A device running Android 5.0 (API 21) or higher is required to run the demo due to the use of the camera2 API, although the native libraries themselves can run on API >= 14 devices.
Other samples are deleted.
SegmentationActivity extends CameraActivity.
This Activity is based on DetectorActivity of previous TF Sample App.
If you just want the fastest path to trying the demo, you may download the nightly build here. Expand the "View" and then the "out" folders under "Last Successful Artifacts" to find tensorflow_demo.apk and libtensorflow_demo.so file.
libtensorflow_inference.so should be re-compiled due to the tf.slice operation compatibility error of nightly-android build.
libtensorflow_inference.so is already compiled in the library folder.
See libs/armeabi-v7a for more details.
Libraries for arm64-v8a device added. libs/arm64-v8a
Once the app is installed it can be started via the "TF Segmentation" which have the orange TensorFlow logo as their icon.
While running the activities, pressing the volume keys on your device will toggle debug visualizations on/off, rendering additional info to the screen that may be useful for development purposes.
- crop size : 129 x 129 :
- crop size : 257 x 257 :
W. Kim and J. Seok, "Indoor Semantic Segmentation for Robot Navigating on Mobile," 2018 Tenth International Conference on Ubiquitous and Future Networks (ICUFN), Prague, 2018, pp. 22-25. [paper-link] [pdf-download]