[InnoWing SRA Training programme 01/2026]

Introduction

Computer vision and deep learning technologies are commonly used in many InnoWing projects. These technologies allow the computers to see and make decisions, either in a digital world or a physical one robotics). Therefore, it is good to experience and learn the recent development in this field.

On-going projects

In the coming semester, InnoWing will host a few projects related to computer vision and deep learning. There are

  1. Crack detection for artificial slope maintenance.
    1. Key problem: How to interact with AI models to enhance the data labeling efficiency for CEDD GEO?
    2. Project lead: Aleks
  2. Denoising for 3D scans of underwater structures.
    1. Key problem: How to develop an effective denoising tool for the scanning team of CEDD survey division?
    2. Project Lead: Ray
  3. 3D reconstruction of InnoWing (or MTR stations) for virtual tour (or maintenance).
    1. Key problem: How to make use of 3D Gaussian splatting to build a high-quality scene for inspection and maintenance purposes?
    2. Project lead: Gene

First milestone:

  1. Report: Assignments are highlighted in the yellow, and a report should be submitted/presented for these questions. The report should contain some examples (videos) to record what you have tried with these tools to contextualize your reporting results.
  2. Submission date & time: The report is expected to be submitted by 23:59pm on Jan 15th and the submission method will be announced later.
  3. Report presentation: We will have a gathering meeting on Jan 16th to share your progress.

Further arrangement and new challenges

The following tasks are very basic and only useful when there is a practical scenario for application. Therefore, if you could finish every task below earlier, please let me know so we can discuss your thoughts or problems encountered. Upon finishing the following tasks, new and more practical problems will be presented to you for a higher level of challenges to really make use of the technologies presented in this document.

I. Experience the power of deep learning-based computer vision

Please try all the tools provided below.