AIM: An Auto-Augmenter for Images and Meshes

Vinit Veerendraveer Singh, Chandra Kambhamettu

CVPR 2022 pp. 722-731

doi:10.1109/CVPR52688.2022.00080 /cvpr/2022/singh2022cvpr-aim/

Abstract

Data augmentations are commonly used to increase the robustness of deep neural networks. In most contemporary research, the networks do not decide the augmentations; they are task-agnostic, and grid search determines their magnitudes. Furthermore, augmentations applicable to lower-dimensional data do not easily extend to higher-dimensional data and vice versa. This paper presents an auto-augmenter for images and meshes (AIM) that easily incorporates into neural networks at training and inference times. It jointly optimizes with the network to produce constrained, non-rigid deformations in the data. AIM predicts sample-aware deformations suited for a task, and our experiments confirm its effectiveness with various networks.

PDF CVPR Semantic Scholar

Cite

Text

Singh and Kambhamettu. "AIM: An Auto-Augmenter for Images and Meshes." Conference on Computer Vision and Pattern Recognition, 2022. doi:10.1109/CVPR52688.2022.00080

Markdown

[Singh and Kambhamettu. "AIM: An Auto-Augmenter for Images and Meshes." Conference on Computer Vision and Pattern Recognition, 2022.](https://mlanthology.org/cvpr/2022/singh2022cvpr-aim/) doi:10.1109/CVPR52688.2022.00080

BibTeX

@inproceedings{singh2022cvpr-aim,
  title     = {{AIM: An Auto-Augmenter for Images and Meshes}},
  author    = {Singh, Vinit Veerendraveer and Kambhamettu, Chandra},
  booktitle = {Conference on Computer Vision and Pattern Recognition},
  year      = {2022},
  pages     = {722-731},
  doi       = {10.1109/CVPR52688.2022.00080},
  url       = {https://mlanthology.org/cvpr/2022/singh2022cvpr-aim/}
}