Try different models for better fps

[Bill-Haoyu-Lin]

Try different yolov5-tiny other model with small dataset, since we need a smaller model for NvidiaTX2 developer board.

• [ ] Yolov3-tiny
• [x] Yolov5-tiny
• [ ] Yolov7-tiny

[Bill-Haoyu-Lin]

Used yolov5s for smaller model try. Fps increase but classifier need to be improved for yolov5

[Bill-Haoyu-Lin]

tried yolov5n for 30 + fps in stable condition and max of 58 fps,

[Bill-Haoyu-Lin]

Need to train model to fit on more various data ( Armor )

• Different lighting condition, with different exposure level
• Different Robot Posture yaw/pitch/roll
• Different camera, we are observing a difference between cameras

We will also need to input the data for labels ( letters around the battle ground ),

We will still be using tiny model for this task, may move to custom model without yolo suite