Methods › Computer Vision › Vision Transformers › MoCo v3

MoCo v3

10 papers tagged archive 2025-07-28

Introduced by Xinlei Chen et al. in An Empirical Study of Training Self-Supervised Vision Transformers

archive 2025-07-28 Description, source and code snippet are the archive's method entry.

MoCo v3 aims to stabilize training of self-supervised ViTs. MoCo v3 is an incremental improvement of MoCo v1/2. Two crops are used for each image under random data augmentation. They are encoded by two encoders f_q and fₖ with output vectors q and k. q behaves like a "query", where the goal of learning is to retrieve the corresponding "key". The objective is to minimize a contrastive loss function of the following form:

ℒ_𝓆=-log(exp(q ·k⁺ / τ))/(exp(q ·k⁺ / τ)+∑_(k⁻) exp(q ·k⁻ / τ))

This approach aims to train the Transformer in the contrastive/Siamese paradigm. The encoder f_q consists of a backbone (e.g., ResNet and ViT), a projection head, and an extra prediction head. The encoder fₖ has the back the backbone and projection head but not the prediction head. fₖ is updated by the moving average of f_q, excluding the prediction head.

PaperSource

Papers archive 2025-07-28

10 shown of 10, newest first. Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code; it is per sample and not a correctness claim.

Tasks archive 2025-07-28

20 shown of 42 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.

TaskPapers
Self-Supervised Learning6
Contrastive Learning3
Self-Supervised Image Classification3
Semantic Segmentation3
Image Classification2
Linear evaluation2
Representation Learning2
image-classification2
Atari Games1
Atari Games 100k1
Classification1
Data Augmentation1
Deep Reinforcement Learning1
GPU1
Image Augmentation1
Image Captioning1
Image Generation1
Image Segmentation1
Knowledge Distillation1
Language Modeling1

Usage over time archive 2025-07-28

Papers per year tagged with MoCo v3: 2021 to 2025, peak 3 3 0 2021: 2 papers 2021 2022: 3 papers 2022 2023: 2 papers 2023 2024: 2 papers 2024 2025: 1 paper 2025
Papers per year the archive tags with this method, by the paper's archive date (10 dated). Bars are counts, not a trend claim.

Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).

Categories archive 2025-07-28

Vision Transformers

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections