ScaleNet: An Unsupervised Representation Learning Method for Limited Information

TL;DR

ScaleNet is an unsupervised learning method that uses multi-scale images and rotation prediction to improve CNN performance with limited data, outperforming previous methods like RotNet on CIFAR-10 and ImageNet.

Contribution

The paper introduces ScaleNet, a novel unsupervised learning approach leveraging multi-scale images and rotation tasks to enhance CNN feature learning with limited data.

Findings

ScaleNet outperforms RotNet by ~7% on CIFAR-10.

Transferred parameters from ScaleNet improve ImageNet classification by ~6%.

ScaleNet enhances other models like SimCLR by learning effective features.

Abstract

Although large-scale labeled data are essential for deep convolutional neural networks (ConvNets) to learn high-level semantic visual representations, it is time-consuming and impractical to collect and annotate large-scale datasets. A simple and efficient unsupervised representation learning method named ScaleNet based on multi-scale images is proposed in this study to enhance the performance of ConvNets when limited information is available. The input images are first resized to a smaller size and fed to the ConvNet to recognize the rotation degree. Next, the ConvNet learns the rotation-prediction task for the original size images based on the parameters transferred from the previous model. The CIFAR-10 and ImageNet datasets are examined on different architectures such as AlexNet and ResNet50 in this study. The current study demonstrates that specific image features, such as Harris corner information, play a critical role in the efficiency of the rotation-prediction task. The ScaleNet supersedes the RotNet by ~7% in the limited CIFAR-10 dataset. The transferred parameters from a ScaleNet model with limited data improve the ImageNet Classification task by about 6% compared to the RotNet model. This study shows the capability of the ScaleNet method to improve other cutting-edge models such as SimCLR by learning effective features for classification tasks.