Any thoughts on using a co-processor for game piece detections for the 2025 reefscape season?
Since we’re technically not picking up any game pieces from the ground like last year’s season (Crescendo). The only possible case is to pick up corals from the loading zone.
Another possibility is the time cuts on cycles if you intake from the floor, so being able to accurately assess the orientation of coral may come in handy. Similar to the 2022 season, you can lock onto the algae to make it easier to pick up. I think there is definitely a high potential for vision this year in both object detection and AprilTag detection (22 AprilTags).
I’d say that although Algae isn’t coming from a feeder, they are still within Known positions on the reef regardless (correct me if I’m wrong). I believe it’d be a waste of resources to have object detection just for algae.
It could be worth looking into it for coral, though.
From the get go object detection is worth it only if it’s a task the robot would have to repeat, or in places that would be difficult for the driver to do themselves. When it comes to algae, if you’re not going to tackle it as a mission you’re going to do throughout the game, and only in your half of the field, I don’t think object detection would help you much. But if you’re planning to try and score many algae during the game, it would also be beneficial to be able to intake from impromptu positions.
I think because it’s a big colored ball, object detection for algae is feasible, but much less so for coral, which is a white cylinder. Cylinders are hard, because it’s hard to see the orientation precisely. Besides, white is not a distinctive color. I think that since it’s easy to intake coral from the coral station there’s not much point to ground intake and vision.