Unit 2: YOLO chapter - Outline for the section

[AntonisCSt]

Hey guys,

These are the proposed sections for the Yolo Section.

Object Detection

• Connection with the previous section.
• How it is different from localization and classification?
• Sliding window approach (theoretically)
• R-CNN, Fast R-CNN (again theoretically not too many details)
• Outro → Methods still slow

Yolo

• What it is and other introductory things about YOLO -Mentioning Pros (Fast and very good mean average precision(mAP) compared to other state-of-the-art object detection) paper [1]

Architecture

• Yolo difference with the previous method (SxS grid, bounding boxes) and CNN architecture
• Minmax expression and more details for the initial steps of Yolo
• Reference to Unit’s 6 how to evaluate object detection (link)
• Yolo Architecture Intuition and what makes it so fast. -Loss function

Yolo over the years

How did it evolve? Where is it today? Main differences

Using Yolov8

The code part goes here to how it is used. (Maybe we could add other cool stuff about Yolo here if it does not make it complicated)

Tasks

• [x] #21
• [ ] #22
• [x] #23
• [ ] #24

References

Open to suggestions for further improvement of the draft!

[sezan92]

Object Detection

• Connection with the previous section.
• How it is different from localization and classification?
• Sliding window approach (theoretically)
• R-CNN, Fast R-CNN (again theoretically not too many details)
• Outro → Methods still slow

I think case of "Connection with the previous section", we can write "Connection with Convolutional Neural Network"

"How it is different from localization and classification -> how it is different from image classification"

"Previous methods,

• sliding window
• RCNN , Fast RCNN, Faster RCNN,
• "

[sezan92]

Architecture

• Yolo difference with the previous method (SxS grid, bounding boxes) and CNN architecture
• Minmax expression and more details for the initial steps of Yolo
• Reference to Unit’s 6 how to evaluate object detection (link)
• Yolo Architecture Intuition and what makes it so fast. -Loss function

I think following order is better,

"

1. Yolo , difference with the previous methods,
2. YOLO architecture intuition
3. Innovations -> (a) Loss function , (b) min-max expression
4. Evaluation ,

[sezan92]

Yolo over the years

How did it evolve? Where is it today? Main differences

Let us describe , differences/ advancements among Yolov1, Yolov2, YOLOV3, ....Yolov8

Using Yolov8

The code part goes here to how it is used. (Maybe we could add other cool stuff about Yolo here if it does not make it complicated)

For first PR I prefer writing how to use existing Yolov8 library , rather than building the architecture. it will make the PR faster, and we get feedbacks earlier.

[themurtazanazir]

How deep are we planning to go into different architectures? I think we should explain YOLOv1 pretty good and then write changes from previous versions for v2-v8

[sezan92]

How deep are we planning to go into different architectures? I think we should explain YOLOv1 pretty good and then write changes from previous versions for v2-v8

I agree