Deep Learning model to develope Simulink/Android app

I'm using the Simulink® Support Package for Android™ Devices to implement a pre-trained DAGNetwork that performs semantic segmentation on images captured by the Android camera.

The model consists of: Android camera block - Concatenate - Predict - function block (to convert single signal (m x n x 3) to RGB) - Video display.

If I consider as the first block "Image From File" to run the model on the PC, it works, but once the previous model is built and deployed on the Android device, the display in the built-in app is black. I think it appears because in this case the input to the predict block is no longer a static image but a "dynamic" signal.

Thanks in advance for any kind of suggestion that solves the problem.

Answer 1

[Sutanu Maiti]

You may have a look at the below example:

https://www.mathworks.com/help/supportpkg/android/ref/detect-and-track-objects-using-deep-learning-on-android-device.html

You may refer the way image taken from camera and converted to an image from the example model:

>> open_system('androidObjectClassification')

The blank screen could be due to the output image after prediction that was given to the cideo display block. You may take input image and predicted label and use any insert text function to embed the predicted text into the input image and then feed that to the video display block.

Hope it helps.