High accuracy and low loss in training but low accuracy and high loss when validation

calmiLovesAI / Basic_CNNs_TensorFlow2

A tensorflow2 implementation of some basic CNNs(MobileNetV1/V2/V3, EfficientNet, ResNeXt, InceptionV4, InceptionResNetV1/V2, SENet, SqueezeNet, DenseNet, ShuffleNetV2, ResNet).

MIT License

524 stars 178 forks source link

High accuracy and low loss in training but low accuracy and high loss when validation #36

Closed s2244521 closed 1 year ago

s2244521 commented 1 year ago

My validation data is the same as training data but the validation result is not good as training in training process. Anyone met this issue?

s2244521 commented 1 year ago

I think BN, dropout adn augmentation are the reasons.

s2244521 commented 1 year ago

solved