Piggyback: Adapting a Single Network to Multiple Tasks by Learning to Mask Weights

Benchmark Model Rank Results
continual-learning-on-imagenet-fine-grained-6-tasksPiggyback#2Accuracy: 76.16
continual-learning-on-sketch-fine-grained-6-tasksPiggyback#3Accuracy: 79.91
continual-learning-on-cubs-fine-grained-6-tasksPiggyback#4Accuracy: 80.5
continual-learning-on-flowers-fine-grained-6-tasksPiggyback#4Accuracy: 94.77
continual-learning-on-stanford-cars-fine-grained-6-tasksPiggyback#4Accuracy: 89.62
continual-learning-on-wikiart-fine-grained-6-tasksPiggyback#5Accuracy: 71.33
continual-learning-on-visual-domain-decathlon-10-tasksPiggyback#7decathlon discipline (Score): 2838Avg. Accuracy: 76.60