StanfordVL/taskonomy

Taskonomy: Disentangling Task Transfer Learning [Best Paper, CVPR2018]

What it solves

Taskonomy addresses the challenge of transfer learning in computer vision, specifically how to determine which pretrained models (tasks) are most useful for a new, unseen task. It aims to disentangle task transfer learning by analyzing the relationships and affinities between different vision tasks.

How it works

The project provides a "task bank" of pretrained models and a corresponding dataset. By analyzing task affinities, the project allows researchers to understand how knowledge from one vision task helps in learning another, measuring the raw and normalized data of these relationships.

Who it’s for

Computer vision researchers and developers who want to leverage transfer learning and understand the relationship between different visual perception tasks.

Highlights

  • Pretrained models available in both PyTorch and TensorFlow.
  • Includes a comprehensive dataset and reference code for task affinity analysis.
  • Based on a CVPR 2018 Best Paper Award winning research paper.

Related

  • Project
    aqlaboratory/openfold-3

    An open-source reproduction of AlphaFold3 for predicting the 3D structures of proteins, RNA, DNA, and small molecules.

  • Project
    freesurfer/freesurfer

    FreeSurfer is an open-source software suite for processing, analyzing, and visualizing human brain MRI data from structural and functional studies.

  • Project
    pynicolas/FairScan

    An open-source Android document scanner that uses local machine learning for automatic document detection and perspective correction to create private PDFs.

  • Project
    SegmentationBLWX/sssegmentation

    A PyTorch-based supervised semantic segmentation toolbox providing a unified framework and extensive model zoo for training and testing segmentation algorithms.