Taskonomy: Disentangling Task Transfer Learning

This repository contains:

• pretrained models (task bank) [PyTorch + TensorFlow].
• dataset
• reference code
• task affinity analyses and results

for the the following paper:

Taskonomy: Disentangling Task Transfer Learning (CVPR 2018, Best Paper Award)

Amir Zamir, Alexander Sax*, William Shen*, Leonidas Guibas, Jitendra Malik, Silvio Savarese.

TASK BANK	DATASET
The taskbank folder contains information about our pretrained models, and scripts to download them. There are sample outputs, and links to live demos.	The data folder contains information and statistics about the dataset, some sample data, and instructions for how to download the full dataset.

Task affinity analyses and results	Website
This folder contains the raw and normalized data used for measuring task affinities.	The webpage of the project with links to assets and demos.

Citation

If you find the code, models, or data useful, please cite this paper:

@inproceedings{zamir2018taskonomy,
title={Taskonomy: Disentangling Task Transfer Learning},
author={Zamir, Amir R and Sax, Alexander and and Shen, William B and Guibas, Leonidas and Malik, Jitendra and Savarese, Silvio},
booktitle={2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
year={2018},
organization={IEEE}
}