Meeting date: 19-04-2023 (10h30)

Summary

This week I improved the dissertation skeleton a little bit, then I decided start with the SOTA, since I was really confused last week about whether yolov5-segmentation is multi-task or not and doing more research it could be that I understand the topic better.

My goal is to show the following topics tomorrow (except red ones):

• (The chapter introduction I will do after finishing the chapter or while I do the deep learning models section)
• Perception - kinds of computer vision algorithms to extract information from an image and explore what is used in the dissertation
  • Image segmentation - Kinds of deep learning based segmentation and what exist for each kind
  • Object detection (2D) -
• Neural networks and deep learning - What are the differences?
  • Single and multi-task neural networks - What are the differences? Why are multi-task networks promising? Why not all multi-loss networks are multi-tasked? Multi-task “levels” and “sub-tasks”
• (Deep learning models - I was thinking in doing this topic and create more “inference modules” at the same time)
• Evaluation metrics - Only topics + Can I include ground truth datasets here?

I didn’t write the problem description yet. I thought about doing this after the research to understand my issue with YOLO, but I ended up starting at the state of the art first and forgot to make this topic.

Next week (27-04)