Obstructy is an academic project which aims to provide a mobile application for basic video processing/editing. Its key purpose is to make available the "Learning to See Through Obstructions" work by Yu-Lun Liu1, Wei-Sheng Lai, Ming-Hsuan Yang, Yung-Yu Chuang, and Jia-Bin Huang [https://alex04072000.github.io/ObstructionRemoval/].
As tools and technologies, the project uses primarily Java, Android Studio (for the mobile app), TensorFlow, Python (for the model), and other tools/libraries for various features that augment the project.
- Android Studio
- Java SDK [recent most LTS would suffice, but any SDK should also work]
- Python 3
- TensorFlow 1.10.0
To utilise the features Obstructy has to offer, you will have to follow the following steps in the provided manner:
- Download the zipped project folder from GitHub.
- Unzip the folder wherever conveniently accessible.
- Open Android Studio.
- Click on ‘File’ and select ‘Open’ from the drop-down menu.
- Go to the required file path and select the unzipped folder.
- Wait a while for the Gradle Building to finish.
- Take an android phone and connect it to your PC via USB cable.
- Turn developer mode on in your android device.
- Enable USB debugging from settings on your mobile.
- Your device will now appear in the list of devices on Android Studio. Select it and press the ‘Run’ button.
- Now the application will run on your mobile.
- From the two options provided on the main screen, tap on either ‘Capture Video’ or ‘Select Video’.
- Once done with either, you will be brought to a main screen where you can preview your video, edit it, share it, and send it to the server for ‘Cleaning’ purposes.
Original (LEFT); Raw Background (CENTER); Raw Foreground (RIGHT)
Original (LEFT); Raw Background (CENTER); Raw Foreground (RIGHT)
Please reach out
