@maxverpoopen

Dump of a lot of things

Sup y’all! First devlog for this project.

I’m planning on making a website that will help with FRC scouting, where basically the website will take a FRC match video, and generate a choreo trajectory from it, from which people can use it for a bunch of statistical data, like determining exact cycle times, can be used by drivers to see how they can improve their skills, and for games where auto strategy is super important (like what happened in Norcal District Champs).

So far, I’ve worked on a wide variety of random things, of which I’ll explain from here:

• Homography function: my plan is to determine the homography transform of the keypoints of the field, and transform it to the actual field, where you can map the pixels of the image to the field position, sooo yea.

• experimentation with OpenCV tracking: experimented with cv tracking functions to focus on a robot, and find the center of the bounding box, from which it can be logged

• making a robot bumper keypoint dataset: creating a dataset on roboflow in which it finds out the 4 keypoints of the robot bumpers, to try and extractas much accuracy, from which it can go through the homography function

challenges:

• estimations of robot pose when its not on the screen

• tracking through missed detections

• inferring robot motions

• temporal consistency (idk what that means mr. gpt)

for now I’ll be working on making the dataset for the robot keypoints, but i think this is a good start!

Humble beginnings

!! I would like to say that this is NOT an original project !! - I’m following a YT tutorial about some person doing this exact same thing, title is same as the project name, name is “Code in a jiffy” or smth

So I’m about an hour through the 5 hour video, and in that time I’ve done a few things…

1. Taken a tennis ball dataset online, and trained YOLO on it

2. Trained keypoints of the tennis court (eg. intersection b/w lines) for future homography calculations

3. Starting to work on the analysis system (basically the end goal of this project is to try and export a trajectory of the tennis ball as it travels on the screen, and overlay its movements on a mini-map on the screen)

I’m using this project to learn more about computer vision and tools in the field, particularly to help my TrajTracker project.

final devlog?

my devlogs so far have been, lackluster to say the least, but it’s now time to explain!

To give some context, im a part of frc team 4550, and one thing that we do every year is to hold a preseason competition in november or so, in which we develop our very own game. We split our team into two separate teams, working to develop their own solutions to this novel game, and invite other local teams as well to compete, so newer members can get experience with making robots overall.

As a senior member, we get the chance to make the game ourselves, so here is a potential option to it.

PRESENTING: CAFFEINE CRASHOUT

Basically the goal of the game is to take these balls (fuel from the 2026 game, called “beans”) and deliver them into these scoring grids (so essentially charged up), but the major catch is that you kinda have to reach up and over the field wall to kind of score into these scoring areas, where the closer the scoring platform is, the lesser points it is. From the image, there are 3 L1 areas, 6 L2 areas, and 9 L3 areas. The holding limit is 3 balls (as accordance to the multiple game piece limit FIRST put out for 2027), and since the balls are 2026 fuel, the hope is for me is that it promotes teams to try and develop something that reaches over and goes down, since fuel is relatively bouncy to shoot or drop from above.

In addition, there are human player stations, where you can get fuel from, there are some fuel on the ground that you can use for auto, etc.

I’m aware that this isn’t like good game design, but what I want is to try and promote some more complex designs from freshman members, which can hopefully promote better designs and complexity from the build season starting in January.

One problem I have with this is that I have zero clue what an endgame structure should be, that could go in the middle of the field, as it is quite empty right now. Another thing I’m considering is whether I should add apriltags, considering the complexity it needs to set something up for this specific field layout, mess with the AprilTagFieldLayout json, blah blah blah.

Doubt anyone will see this, but if you do, plz give me some ideas for the endgame

(also ik my cad is terrible im strategy/programming)

Working on scoring elements and grid

Field, barriers, and human player station