[Paper Review] Rotation(2018), Unsupervised Representation Learning by Pre-diction Image Rotations

Unsupervised Representation Learning by Pre-diction Image Rotations

Spyros Gidaris, Praveer Singh, Nikos Komodakis, arXiv 2018

 

PDF, SSL By SeonghoonYu August 4th, 2021

 

Summary

 

 The ConvNet is trained on the 4-way image classification task of recognizing one of the four image rotation(0, 90, 180, 270). The task of predicting rotation transformations provides a powerful surrogate supervision signel for feature learning and leads to dramatic improvments on the relevant benchmarks. 

 

 

 The ConvNet model must learn to solve this loss function.

 

 

 

 K is a number of discrete geometric transformation. G is geometric transformation. F is the ConvNet. $\theta$ is the parameters of the ConvNet.

Experiment

 Compare the attention maps of RotationNet with the attention maps generated by a model trained on the object recognition task in a supervised way. We observe that both models seem to focus on rougly the same image regions.

 

 Abalation study on the number of recognized rotations.

 

Comparison with other methods

 

 

What I like about the paper

• Simple way to learn useful features for downstream tasks on unsupervised fashion.
• The task of recogniziong rotation transformation has a good performance compared with supervised model. 

 

---

my github about what i read

Seonghoon-Yu/Paper_Review_and_Implementation_in_PyTorch