This project extends version 8.3.70 of Ultralytics by incorporating support for RGB+IR (visible light + infrared) dual-modal input for object detection tasks. Additionally, it provides an adjusted and adapted version of the LLVIP dataset, named LLVIP-For-Ultralytics, for training and testing. Users are encouraged to cite the original work if utilized in research or applications.
Note: Dual-modal input training requires significantly more memory. If the training is automatically terminated with a "Killed" prompt, it may indicate a memory overflow. Monitor memory usage during training and adjust hyperparameters based on the device’s specifications.