Unsupervised HDR Image and Video Tone Mapping via Contrastive Learning
This project presents a unified framework (IVTMNet) for unsupervised image and video tone mapping, utilizing contrastive learning to improve the training process. It introduces a novel latent code to measure similarity between tone-mapped results and employs spatial-feature-enhanced and temporal-feature-replaced modules to enhance information exchange and temporal consistency. The project also provides a large-scale unpaired HDR-LDR video dataset, making it a valuable resource for developers working on HDR video tone mapping and related applications.
Visit the Resource







