/Implementation-of-Learning-without-Forgetting-for-multi-class

Primary LanguagePython

PyTorch Implementation of Learning without Forgetting for multi-class

A PyTorch Implementation of Learning without Forgetting.

The LwF Implement for multi-class

About LwF.MC,you can read iCaRL: Incremental Classifier and Representation Learning

requirement

python3.6

Pytorch1.3.0 linux

PIL

run

Create “model” directory

python -u main.py

Result

Resnet18+CIFAR100（Incomplete fit）

incremental step 0 1 2 3 4 5 6 7 8 9 average
test accuracy 80.06 59.15 53.334 47.778 42.196 38.82 36.592 32.996 30.442 27.97 40.591