CrossEdit: Cross-Modal Training Enables Rich Audio-Visual Editing

Benchmark Model Rank Results
image-editing-on-imgedit-dataCrossEdit–Overall: 3.72Add: 4.20Adjust: 4.29Extract: 1.74Replace: 3.61…
image-editing-on-imgedit-dataCrossEdit-Base–Overall: 3.69Add: 4.37Adjust: 4.06Extract: 1.48Replace: 3.61…