TriSense-2M is a large-scale multimodal dataset for training the TriSense model. It consists of raw data after judger evaluation and three stages of processed training data. We have done another round of filtering compared to the paper version.

Dataset Sources

Repository: https://github.com/zinuoli/TriSense Paper: https://arxiv.org/pdf/2505.18110?

	Uses




	
	Direct Use

This dataset is intended for training and fine-tuning… See the full description on the dataset page: https://huggingface.co/datasets/zinuoli/TriSense-2M.