Hello. The rule about terms and definitions of the robot, especificially talking about topic 4.1.3, states, “Tools which are not developed by the team, which can independently complete a task, or a part of a task, which enables the robot to earn points by sending a signal to the controller without further development (e.g., OCR libraries, pre-trained AI model to identify objects) are prohibited.”
My team uses YOLO object detection to identify victims, is it still allowed?
Hi @menenski!
YOLO can be used as a framework as long as your team is actively developing it by training or fine-tuning it on a custom dataset. The key factor in compliance with rule 4.1.3 is that the model should not be a pre-trained, plug-and-play solution that independently identifies objects without further development.
Csaba
2025 Committee
1 Like
Hi @menenski!
I would like to clarify my previous response. Fine-tuning a pre-trained model does not qualify as development, and therefore does not comply with the rules. However, if the team applies significant modifications to the pre-trained model before fine-tuning (chaning the model architecture, layers, loss function), it is considered compliant with the rules, since besides fine-tuning, you make developments on the model.
Csaba
2025 Committee