T-PhISH-Net Trained Weights for Temporally Consistent Underwater Image Enhancement

This dataset contains the trained model weights used in the paper “Temporally Consistent Underwater Image Enhancement via a Transformer-based Extension of PhISH-Net”.Included models:- tphishnet_final.pth- tphishnet_no_backscatter.pth- tphishnet_learnable_noback.pth- tphishnet_no_temporal.pth- tphishnet_no_transformer.pth- phishnet_base.pth- phishnet_nobackscatter.pthThese weights correspond to all models developed by the author, including ablations.Baseline model weights are not redistributed and must be obtained from the original authorsThe weights were trained using publicly available datasets (UIEB, UVE-38K, Brackish, MVK, UOT32, VDD-C).No new proprietary data were created.The corresponding code for running and evaluating the models is available at:https://github.com/bcwein/UIE_benchmark