Georgia Tech’s ZipIt! Effectively Merges Vision Models Trained on Disjoint Tasks Without Additional Training

In the new paper ZipIt! Merging Models from Different Tasks Without Training, a Georgia Tech research team proposes ZipIt!, a general method that exploits redundant features to combine two or more models with the same architecture but trained on different tasks into one multi-task model without additional training.

ai architecture artificial intelligence computer vision & graphics deep-neural-networks exploits features general georgia georgia tech machine learning machine learning & data science merging ml multitask learning paper research research team team tech technology training vision vision models