COMPUTER VISION CAPABILITIES FOR AQUATIC NAVIGATION

Software vision team:

• Salvador Cantuarias
• Nouhaila Elmalouli

Tasks:

2025-04-11:

Create a class that uses computer vision's color recognition to detect elements in the video feed and make the boat respond accordingly.
status: DONE

observations:

• The script is able to take pre recorded video inputs. now its time to test live feed.
  
• We need to change the mathematical approach to identifing the postion of the buoys. Right now it is just related to which forizontal half of the frame is being used. this ould be changed by measuring the delta space between the buoys, if said delta space has a postive or negative value and how close is the center of that delta space to the center of the frame for color correction.
  
• More cases need to be tested: when theres more than two buoys, when a buoy is partially obstrucuted, etc...
  
• Cardinal buoys need to be integrated into the class.

2025-04-22:

Detect red and green bouys + give directions according to different dispositions cases
status: DONE
assigned to: Nouhaila

reworking the mathematical system for making desisions (delta distance between buoys)
status: DONE
assigned to: Salvador

Observations:

generate or record more cases of buoys position.
status: DONE
assigned to: Salvador

Observations:
The race path was reconstucted in rhinoceros to the exact specifications given by the competitons handbook and a video was recorded to be used as input. This new footage shed more light on what are the elements to take into consideration and where the script's weakpoints are.
A video of the current performance (2025-04-20) is provided

add cardinal buoys.
status: DONE
assigned to: Nouhaila

Observations:

tap into real time video feed.
status: STARTED
assigned to: Salvador

Observations:

get some ideas of how to detect the OTTER (unknown obstacle).
status: NOT STARTED
assigned to: Nouhaila, Salvador

Observations:

2025-04-29:

use the given dataset and try object detection to get better segmentation of the buoys.
status: DONE
assigned to: Salvador

Try different YOLO models + Train it for object segmentation (bouys/obstacles/background elements...).
status: STARTED
assigned to: Nouhaila

Observations:
link to new updates in performance