Abstract
Recent self-supervised representation learning techniques have largely closed the gap between supervised and unsupervised learning on ImageNet classification. While the particulars of pretraining on ImageNet are now relatively well understood, the field still lacks widely accepted best practices for replicating this success on other datasets. As a first step in this direction, we study contrastive self-supervised learning on four diverse large-scale datasets. By looking through the lenses of data quantity, data domain, data quality, and task granularity, we provide new insights into the necessary conditions for successful self-supervised learning. Our key findings include observations such as: (i) the benefit of additional pretraining data beyond 500k images is modest, (ii) adding pretraining images from another domain does not lead to more general representations, (iii) corrupted pretraining images have a disparate impact on supervised and self-supervised pretraining, and (iv) contrastive learning lags far behind supervised learning on finegrained visual classification tasks.
Original language | English |
---|---|
Title of host publication | 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) |
Publisher | Institute of Electrical and Electronics Engineers |
Pages | 14735-14744 |
Number of pages | 10 |
ISBN (Electronic) | 978-1-6654-6946-3 |
ISBN (Print) | 978-1-6654-6947-0 |
DOIs | |
Publication status | Published - 27 Sept 2022 |
Event | IEEE/CVF Conference on Computer Vision and Pattern Recognition 2022 - New Orleans, United States Duration: 19 Jun 2022 → 24 Jun 2022 https://cvpr2022.thecvf.com/ |
Publication series
Name | Conference on Computer Vision and Pattern Recognition (CVPR) |
---|---|
Publisher | IEEE |
ISSN (Print) | 1063-6919 |
ISSN (Electronic) | 2575-7075 |
Conference
Conference | IEEE/CVF Conference on Computer Vision and Pattern Recognition 2022 |
---|---|
Abbreviated title | CVPR 2022 |
Country/Territory | United States |
City | New Orleans |
Period | 19/06/22 → 24/06/22 |
Internet address |