EfficientNetV2: Smaller Models and Faster Training

TL;DR

EfficientNetV2 introduces a family of convolutional networks that train faster, are more parameter-efficient, and outperform previous models on image classification benchmarks, achieved through neural architecture search and progressive learning techniques.

Contribution

The paper presents EfficientNetV2, a new neural network family optimized for faster training and better efficiency, using training-aware architecture search and adaptive regularization.

Findings

EfficientNetV2 trains 5-11x faster than previous models.

Models are up to 6.8x smaller with improved accuracy.

Achieves 87.3% top-1 accuracy on ImageNet, surpassing ViT.

Abstract

This paper introduces EfficientNetV2, a new family of convolutional networks that have faster training speed and better parameter efficiency than previous models. To develop this family of models, we use a combination of training-aware neural architecture search and scaling, to jointly optimize training speed and parameter efficiency. The models were searched from the search space enriched with new ops such as Fused-MBConv. Our experiments show that EfficientNetV2 models train much faster than state-of-the-art models while being up to 6.8x smaller. Our training can be further sped up by progressively increasing the image size during training, but it often causes a drop in accuracy. To compensate for this accuracy drop, we propose to adaptively adjust regularization (e.g., dropout and data augmentation) as well, such that we can achieve both fast training and good accuracy. With progressive learning, our EfficientNetV2 significantly outperforms previous models on ImageNet and CIFAR/Cars/Flowers datasets. By pretraining on the same ImageNet21k, our EfficientNetV2 achieves 87.3% top-1 accuracy on ImageNet ILSVRC2012, outperforming the recent ViT by 2.0% accuracy while training 5x-11x faster using the same computing resources. Code will be available at https://github.com/google/automl/tree/master/efficientnetv2.