Project done in order to learn how to use DeepLearning models in mobile applications. A camera application with real time semantic segmentation was implemented using OpenCV(for image preprocessing and camera handling) and Google´s trained Deeplab3+ model(for video frame segmentation).
![person](Demos/person.gif)
![walking](Demos/walking.gif)
![plants](Demos/plants.gif)
![dog](Demos/dog.gif)
![sofa](Demos/sofa.gif)
![table_chairs](Demos/table_chairs.gif)
- Use of quantized version of deeplab3+(for better inference speed): https://github.com/tantara/JejuNet
- Visualize .tflite model structure/inputs/outputs: https://github.com/lutzroeder/netron
- Fix opencv camera: https://heartbeat.fritz.ai/working-with-the-opencv-camera-for-android-rotating-orienting-and-scaling-c7006c3e1916