YFCC100M

By Bart Thomee, David A. Shamma, Gerald Friedland, Benjamin Elizalde, Karl S. Ni, Douglas N. Poland, Damian Borth, Li-Jia Li. 2015

995 citations recorded by the source corpus.

This publicly available curated dataset of almost 100 million photos and videos is free and legal for all.

Subfields: Cancer Research, Radiology, Nuclear Medicine and Imaging, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cancer Research

1. Han-Jia Ye
2. Martial Hebert
3. Yu-Xiong Wang
4. Ishan Misra
5. Peng Cui
6. Yu Qiao
7. Andrea Vedaldi
8. Ming-Hsuan Yang

In Radiology, Nuclear Medicine and Imaging

1. Zeynep Akata
2. Bernt Schiele
3. Han-Jia Ye
4. Yongqin Xian
5. Trevor Darrell
6. Ishan Misra
7. Andrea Vedaldi
8. Xinlei Chen

In Evaluation & Benchmarks

1. Trevor Darrell
2. Li Fei-Fei
3. Marcus Rohrbach
4. Bernt Schiele
5. Cordelia Schmid
6. Abhinav Kumar Gupta
7. Jianfeng Gao
8. Mohit Bansal

Citations in the corpus

• Learning Transferable Visual Models From Natural Language Supervision (2021, methodology): “They converted the title, description, and hashtag metadata of images in the YFCC100M dataset (Thomee et al., 2016) into a bag-of-words multi-label classification task and showed that pre-training AlexNet (Krizhevsky et al., 2012) to predict these labels learned representations which preformed…”
• Big Transfer (BiT): General Visual Representation Learning (2019, methodology): “In [23,30] the authors use a dataset containing 100M Flickr images [56].”
• Self-Training With Noisy Student Improves ImageNet Classification (2019, neutral): “To enable fair comparisons with our results, we also experiment with a public dataset YFCC100M [85] and show the results in Appendix A.”
• VirTex: Learning Visual Representations from Textual Annotations (2020, methodology): “[40] trains visual N-gram models on the YFCC-100M dataset [41], that provides 100M Flickr images with user-provided tags.”
• Zero-Shot Text-to-Image Generation (2021, methodology): “This dataset does not include MS-COCO, but does include Conceptual Captions and a filtered subset of YFCC100M (Thomee et al., 2016).”
• Self-Supervised Learning of Pretext-Invariant Representations (2019, methodology): “To study how much changes in the data distribution impact PIRL, we pre-train models on uncurated images from the unlabeled YFCC dataset [72].”
• Scaling and Benchmarking Self-Supervised Visual Representation Learning (2019, methodology): “We note that our Jigsaw YFCC-100M self-supervised model outperforms both the supervised models (ImageNet-1k and Places205 supervised) across all the metrics by a significant margin, e.g., a 5 point gain compared to the Places205 supervised model on the number of pixels within t◦=11.5 metric.”
• LiT: Zero-Shot Transfer with Locked-image text Tuning (2021, methodology): “Specifically, we tune an ImageNet-21k pre-trained ViTL/16 model [54] on the union of the YFCC100m-CLIP and CC12M datasets.”
• NExT-QA: Next Phase of Question-Answering to Explaining Temporal Actions (2021, neutral): “To 1Videos are drawn from YFCC-100M [43] and are crawled from Flickr.”
• ELEVATER: A Benchmark and Toolkit for Evaluating Language-Augmented Visual Models (2022, neutral): “(ii) Academic track is a setting that limits the data in established public large datasets (i.e., ImageNet-21K [9], GCC3M [54] & 12M [5], YFCC15M [61]).”
• Unified Contrastive Learning in Image-Text-Label Space (2022, positive): “We first reproduced CLIP on YFCC-14M with Swin-Tiny.”
• Localizing Moments in Video with Natural Language (2017, methodology): “ method is used to align action detection annotations to visual descriptions in Charades. Hendricks et al. (2017) collected the Distinct Describable Moment (DiDeMo) dataset, which consists of Flickr (Thomee et al., 2016) videos with localized referring expressions. Both Charades and DiDeMo contain a”
• In Defense of Grid Features for Visual Question Answering (2020, methodology): “For example, YFCC contains a lot of outdoor images [40], which are not present in the VQA 4https://github.com/MILVLG/mcan-vqa dataset.”
• Meta Pseudo Labels (2020, methodology): “To avoid using proprietary data like JFT, we use the ImageNet training set as labeled data and the YFCC100M dataset [65] as unlabeled data.”
• MultiGrain: a unified image embedding for classes and instances (2019, methodology): “cating the mean average precision on the “strong” subset of the INRIA Copydays dataset [8]. We add 10K distractor images randomly sampled from the YFCC100M large-scale collection of unlabelled images [42]. We call the combination C10k. The PCA whitening transformations are computed from 6 200 400 60”