Ekin D. Cubuk, Barret Zoph, Dandelion Mane, Vijay Vasudevan, Quoc V. Le
Google
Submitted on 24 May 2018
Paper: https://arxiv.org/abs/1805.09501
Code: https://github.com/tensorflow/models/tree/master/research/autoaugment
Comments: First paper in the field, same idea as AutoML: using RL to search the augmentation policy. Potential gains are shown, however, the main problem is the computation burden.
Ekin D. Cubuk, Barret Zoph, Jonathon Shlens, Quoc V. Le
Google
Submitted on 30 Sep 2019
Paper: https://arxiv.org/abs/1909.13719
Code: https://github.com/tensorflow/tpu/blob/master/models/official/efficientnet/autoaugment.py
Comments: The main contribution is the reduced search space. Only two parameters left: N and M. (N: Number of augmentation transformations to apply sequentially; M: Magnitude for all the transformations)
Dan Hendrycks, Norman Mu, Ekin D. Cubuk, Barret Zoph, Justin Gilmer, Balaji Lakshminarayanan
Google
Submitted on 5 Dec 2019
Paper: https://arxiv.org/abs/1912.02781
Code: https://github.com/google-research/augmix
Comments: Less severe augmentation policies and JS diversity loss are applied.
Questions: What's the sepearte effect for less severe policy and JS diversity loss? Is it benificial for clean data rather than corrupted datasets?
Ali Dabouei, Sobhan Soleymani, Fariborz Taherkhani, Nasser M. Nasrabadi
West Virginia University
Submitted on 10 Mar 2020
Papaer: https://arxiv.org/abs/2003.05034
Code: https://github.com/alldbi/SuperMix
Comments: SuperMix, which exploits the knowledge of a teacher to mix images based on their salient regions. SuperMix optimizes a mixing objective that considers: i) forcing the class of input images to appear in the mixed image, ii) preserving the local structure of images, and iii) reducing the risk of suppressing important features.
https://github.com/aleju/imgaug
This python library helps you with augmenting images for your machine learning projects. It converts a set of input images into a new, much larger set of slightly altered images.
TBA
TBA