PODNet: Pooled Outputs Distillation for Small-Tasks Incremental Learning

Benchmark Model Rank Results
incremental-learning-on-imagenet-500-classes-10-steps-of-50-classesPODNet#2Average Incremental Accuracy: 64.13
incremental-learning-on-imagenet-100-50-classes-25-steps-of-2-classesPODNet#3Average Incremental Accuracy: 67.28
incremental-learning-on-imagenet-500-classes-5-steps-of-100-classesPODNet#3Average Incremental Accuracy: 66.95
incremental-learning-on-cifar-100-50-classes-25-steps-of-2-classesPODNet#4Average Incremental Accuracy: 60.72
incremental-learning-on-imagenet-100-50-classes-5-steps-of-10-classesPODNet#4Average Incremental Accuracy: 75.82
incremental-learning-on-imagenet-100-50-classes-10-steps-of-5-classesPODNet#5Average Incremental Accuracy: 73.14
incremental-learning-on-cifar-100-50-classes-10-steps-of-5-classesPODNet (CNN)#9Average Incremental Accuracy: 63.19
incremental-learning-on-cifar-100-50-classes-5-steps-of-10-classesPODNet (CNN)#9Average Incremental Accuracy: 64.83
incremental-learning-on-cifar-100-b0-5steps-of-20-classesPODNet#9Average Incremental Accuracy: 66.70