Momentum Contrast for Unsupervised Visual Representation Learning
2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020 · paper · AI and machine learning
Presents Momentum Contrast (MoCo), which builds a large, consistent dictionary for contrastive unsupervised learning using a queue and a moving-average encoder. Its representations reportedly transfer well, outperforming supervised pre-training on seven detection and segmentation tasks.
Checked: PDF from arxiv.org, fetched 4 Oct 2026. Page numbers are those of that PDF.
0 findings
None that passed a second check.
2 minor (typos and the like; most checked by the judge only)
Typo Sec. 3.1: "or context consisting a set of patches [46]" is missing "of" ("consisting of a set of patches").
- p. 2 Sec. 3.1, last paragraph
or context consisting a set of patches [46]
- p. 2 Sec. 3.1
context consisting a set of patches
Judge's note: The page image (crop C001-a, right column of p.2) reads "context consisting a set of patches [46]". The word "of" is absent from the image as well as the text layer, so this is not an extraction artefact.
Second check: none (judge only)
Report a problem with this finding
Typo Sec. 4.1: "This is inline with our hypothesis" should read "in line with".
- p. 5 Sec. 4.1, memory bank paragraph
This is inline with our hypothesis
Judge's note: The page image (p.5, right column) shows "This is" at the end of one line and "inline with our hypothesis" at the start of the next. "inline" is a single word on the page, not a hyphenation or extraction artefact. This is a cosmetic usage slip.
Second check: none (judge only)