Tasks

Work packages

W1: Character recognition

Given an image, the characters in the image including their position needs to be determined. See e.g. here for a dataset that already does it. We need to train a new neural network for that, as theirs is not usable for mobile. There are pretrained models which we can use for that, found e.g. in the Tensorflow model zoo.

W2: Augmented reality

• Detect characters in the augmented reality world with a neural network and track them
• Collect character 3D models and their decomposition
• Show the 3D models in the scene
• Track the characters if ARCore does not do that
• Denoise it/smooth it

W3: User interface

Detecting characters and displaying them in 3D is not enough. We also want to display useful information about them, e.g. their decomposition, history, etymology, alternative forms e.g. oracle bone, ... . We need to decide what information to show, how to show it to the user. where to get the character information, collect it and then add the showing in the app.

W4: Organisation

• Register with the competition
• Create slides/videos
• ...