ResearchBenchmarksPedestrian Image Captioning on ICFG-PEDES v1.0 (test)Follow25.7B@4Hulk22.37223.23624.124.964Dec 4, 2023Evaluation ResultsMethodMethodLinksB@4CIDErHulkBackbone=ViT-B, Traina...Backbone=ViT-B, Trainable Params=113.7M, Total Params=113.7M2023.1225.757BLIP-2Backbone=ViT-G*, Train...Backbone=ViT-G*, Trainable Params=188.0M, Total Params=1.1B2023.1224.244.2BLIPBackbone=ViT-B, Traina...Backbone=ViT-B, Trainable Params=446.0M, Total Params=446.0M2023.1222.536.9