Abstract
Data augmentation (DA) techniques aim to increase data variability, and thus train deep networks with better generalisation. The pioneering AutoAugment automated the search for optimal DA policies with reinforcement learning. However, AutoAugment is extremely computationally expensive, limiting its wide applicability. Followup works such as Population Based Augmentation (PBA) and Fast AutoAugment improved efficiency, but their optimization speed remains a bottleneck. In this paper, we propose Differentiable Automatic Data Augmentation (DADA) which dramatically reduces the cost. DADA relaxes the discrete DA policy selection to a differentiable optimization problem via Gumbel-Softmax. In addition, we introduce an unbiased gradient estimator, RELAX, leading to an efficient and effective one-pass optimization strategy to learn an efficient and accurate DA policy. We conduct extensive experiments on CIFAR-10, CIFAR-100, SVHN, and ImageNet datasets. Furthermore, we demonstrate the value of Auto DA in pre-training for downstream detection problems. Results show our DADA is at least one order of magnitude faster than the state-of-the-art while achieving very comparable accuracy. The code is available at https://github.com/VDIGPKU/DADA.
Original language | English |
---|---|
Title of host publication | 16th European Conference on Computer Vision, ECCV 2020: proceedings |
Publisher | Springer |
Pages | 580–595 |
Number of pages | 16 |
ISBN (Electronic) | 9783030585426 |
ISBN (Print) | 9783030585419 |
DOIs | |
Publication status | Published - 17 Nov 2020 |
Event | European Conference on Computer Vision 2020 - Glasgow, United Kingdom Duration: 23 Aug 2020 → 28 Aug 2020 https://eccv2020.eu/ |
Publication series
Name | Lecture Notes in Computer Science |
---|---|
Volume | 12367 |
ISSN (Print) | 0302-9743 |
ISSN (Electronic) | 1611-3349 |
Conference
Conference | European Conference on Computer Vision 2020 |
---|---|
Country/Territory | United Kingdom |
City | Glasgow |
Period | 23/08/2020 → 28/08/2020 |
Internet address |