2022-10-13

Landing Pad ML

• Are we provided data for what the obstacles and/or landing pad looks like?

• Do we require really quick performance – do we train the model using more specific data or use the COCO dataset?

• Furthermore, how much do we compromise on accuracy for speed?

• Do we stick YOLOv7 and are there any better training datasets?

• Colour detection – what methods are best applicable?

TO DO:

• Start documenting object detection algorithms and comparisons in Confluence.

• Find good training data sets in addition to COCO.

• Researched a few one-stage models and compared them.

• Focused on speed more than accuracy since we need to detect and avoid obstacles rather than recognize them.

• Found YOLOv7 pretty good for use case.

Autonomous landing (R&D)

Path optimization

Data telemetry

Autonomous landing (comp)

GUI

• Created GUI Starter Project so team members can start designing each page

• Went over project details and the PyQT5 framework

• Assigned tasks

  • Task 1: Motors Page @Amy Hu

  • Task 2: Adding Maps @Harini Karthik

  • Task 3: Attitude Controller @Ethan Woo @Harry Fung

  • Task 4: Displaying Information (Drone info, Setup info, Logging info) @Alex Yang @Manuel Stefan Christopher