This repository is a curated collection of academic papers focused on Self-Supervised Learning (SSL) and Representation Learning, primarily within the Computer Vision domain. It serves as a valuable resource for researchers and practitioners seeking to understand the landscape of SSL techniques, track state-of-the-art performance, and identify key publications.

How It Works

The repository organizes papers by sub-fields within SSL, such as Contrastive Learning, Dense Prediction, and Knowledge Distillation. Each entry typically includes the conference/journal, paper title, a link (implied by arXiv ID or conference), and reported performance metrics (e.g., ImageNet Top-1 accuracy, COCO AP). This structured format allows for quick comparison of different methods and their effectiveness on standard benchmarks.

Quick Start & Requirements

This is a curated list of papers and does not involve code execution. No installation or specific software requirements are needed to browse the content.

Highlighted Details

• Comprehensive coverage of contrastive learning methods, including MoCo, SimCLR, SwAV, and Barlow Twins, with reported ImageNet accuracy.
• Detailed sections on dense prediction tasks, showcasing performance on COCO object detection and segmentation benchmarks.
• Inclusion of papers on self-supervised learning with knowledge distillation, highlighting techniques like SEED and DisCo.
• Broad categorization across various domains including NLP, Speech, Graph, and Reinforcement Learning, though the primary focus remains Computer Vision.

Maintenance & Community

The last update noted was September 26, 2021. Contributions and comments are welcomed, suggesting an open-source, community-driven curation effort.

Licensing & Compatibility

The repository itself, as a collection of links and information, is not subject to typical software licensing. The underlying papers are subject to their respective copyright and licensing terms.

Limitations & Caveats

The repository's content is static as of its last update in September 2021, meaning it does not reflect the rapid advancements in Self-Supervised Learning that have occurred since. It primarily focuses on Computer Vision, with less depth in other domains.