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.
Covered most of the general-purpose benchmarks. Added the best performing architectures from the last years.
- Scene recognition
- Action recognition
- Shape recognition
- Pose estimation
- Object detection
- Image classification
- Semi-supervised learning
- Weakly supervised learning
- Semantic segmentation
- Instance segmentation
- Face recognition
- Face alignment
- Human parsing
- Keypoint and landmark detection
- Domain adaptation
- Image superresolution
- Saliency detection
- Structure from motion
- Image captioning
- Surface reconstruction
- Inverse graphics
- Object localization
- Optical character recognition
- Image representations and feature learning
- Alpha matting
- Adversarial attacks and defences
- Medical imaging
- Image co-segmentation
- Visual tracking
- Visual question answering
- Optical flow estimation
- Depth estimation
- Image retrieval
- Stereo matching
- Image synthesis
- Structure learning
- Image inpainting
- Trajectory prediction
- Image warping
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.