COCO-MLT

Dataset Information
Modalities
Images
Languages
English
Introduced
2020
License
Unknown
Contents

Overview

The COCO-MLT is created from MS COCO-2017, containing 1,909 images from 80 classes. The maximum of training number per class is 1,128 and the minimum is 6. We use the test set of COCO2017 with 5,000 for evaluation. The ratio of head, medium, and tail classes is 22:33:25 in COCO-MLT.

Variants: COCO-MLT

Associated Benchmarks

This dataset is used in 2 benchmarks:

Recent Benchmark Submissions

Task Model Paper Date
Long-tail Learning LMPT(ViT-B/16) LMPT: Prompt Tuning with Class-Specific … 2023-05-08
Long-tail Learning LMPT(ResNet-50) LMPT: Prompt Tuning with Class-Specific … 2023-05-08
Zero-Shot Learning ResNet-50 Learning Transferable Visual Models From … 2021-02-26
Zero-Shot Learning ViT-B/16 Learning Transferable Visual Models From … 2021-02-26
Long-tail Learning CLIP(ViT-B/16) Learning Transferable Visual Models From … 2021-02-26
Long-tail Learning CLIP(ResNet-50) Learning Transferable Visual Models From … 2021-02-26
Long-tail Learning DB Focal(ResNet-50) Distribution-Balanced Loss for Multi-Label Classification … 2020-07-19
Long-tail Learning LDAM(ResNet-50) Learning Imbalanced Datasets with Label-Distribution-Aware … 2019-06-18
Long-tail Learning OLTR(ResNet-50) Large-Scale Long-Tailed Recognition in an … 2019-04-10
Long-tail Learning ML-GCN(ResNet-50) Multi-Label Image Recognition with Graph … 2019-04-07
Long-tail Learning CB Loss(ResNet-50) Class-Balanced Loss Based on Effective … 2019-01-16
Long-tail Learning Focal Loss(ResNet-50) Focal Loss for Dense Object … 2017-08-07
Long-tail Learning RS(ResNet-50) Relay Backpropagation for Effective Learning … 2015-12-18

Research Papers

Recent papers with results on this dataset:

External Links: