sankaushik / sota-cv

A repository of state-of-the-art deep learning methods in computer vision

SotA-CV

A repository of state-of-the-art deep learning results in computer vision. It aims to collect and maintain up-to-date information on the latest developments in in computer vision, facilitating the research effort in deep learning.

Unlike other attempts in collaborative tracking of research progress, this repository provides aggregate results of quantitative evaluation. Such practice allows to greatly simplify both the initial literature search and preparing a comparative study of your own results.

Tasks

Contents

• Semantic segmentation
• Image classification
• Object detection
• Monocular depth estimation
• Pose estimation
• Image Quality Assessment

TODO

• Semi-supervised classification
• Weakly-supervised semantic segmentation
• Scene recognition
• Action recognition
• Shape recognition
• Face recognition
• Face alignment
• Keypoint and landmark detection
• Instance segmentation
• Human parsing
• Saliency detection
• Structure from motion
• Image captioning
• Surface reconstruction
• Inverse graphics
• Object localization
• Optical character recognition
• Image representations and feature learning
• Medical imaging
• Image co-segmentation
• Visual tracking
• Visual question answering
• Optical flow estimation
• Image retrieval
• Stereo matching
• Image synthesis
• Structure learning
• Image inpainting
• Trajectory prediction
• Image warping
• Domain adaptation
• Adversarial attacks and defences

Datasets

Contents

• PASCAL VOC 2012 — Semantic Segmentation
• ADE20K
• Cityscapes — Semantic Segmentation
• PASCAL Context
• COCO — Detection 2016

Contributions

Pull requests are most welcome. To make the material more coherent, please follow the examples in dataset and problem templates.

For your convenience, use incoming papers list. In supplementary docs there's a tutorial on metrics and datasets.