Training with custom dataset

[cskn95]

Hello;

We're a team preparing for autonomous armed UAV competition. What they expect from our UAV is detect and track other UAV's for at least 4 second without a break. We decided to train and use YOLOv8 with our custom dataset. Our plan is to send the start box to Re3 with YOLOv8 and track the enemy but we can not managed to train Re3 for our purpose.

We have approx. 10 thousand train image and a thousand val image for training and same amount of .txt file that only includes 5 values for bounding box. You can find the dataset below for more information end expectations.

We need your help to reshape our dataset for Re3 and train process.

https://drive.google.com/drive/folders/1gFC08NXcVagYdLhH-oLri4vgE3BhpgPz?usp=sharing