In this group assignment (group size 3-4 persons) you will implement a system that uses of a video stream from a camera and recognises and marks some objects in the image using an off-the-shelf deep learning based computer vision module. The requirements are that you have performed at least some minor fine-tuning to the model so that it can recognise some other type of objects and that this fine tuning is not described exactly somewhere already (for example, don't follow a blog post where someone tells you exactly what to do).

What type of objects do you want to recognise?

What models are out there that gets you close?

How to adapt?

How easy to extend?

The project groups will be defined in part based on your prior knowledge with the idea that each group should have at least someone that knows more or less how to solve the deep learning part of such a problem. For this assignment we would also like you do, when possible, work with people at your university so that you get to expand your network locally. Some of you will be from the same group/division in that case please try to pick different groups.

The work is divided into two parts. Please take a look at these before starting the work so that you see what is expected of you.

Assignment Object detection Part 1
- This is completed before the 2-day session in LiU - Sensing and Perception and should be a plan for how to do it without spending too much work. This will then be presented in Linköping and the groups pick the best ideas for Part 2.
Assignment Object detection Part 2
- This is where you actually implement things