CelebA Multi-Task

[hussam789]

Hey, Can you provide an example of how to train multi-task classification such as:

• Visual Domain Decathlon
• CelebA

[senwu]

Thanks for your interests.

We'd love to support more applications. Can you provide more details about these datasets? e.g., description, link.

Also, we have several applications in Emmental Tutorials.

Thanks, Sen

[hussam789]

Thanks for replying, CelebA 200K celebrity images, each with 40 attribute annotations: http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html

Visual Domain Decathlon: consists of 10 different image classification tasks. https://www.robots.ox.ac.uk/~vgg/decathlon/

I've looked at the Tutorial, and I didn't see a tutorial for multi-classification tasks (multi-task).

I think adding a tutorial for one of these common benchmarks will be beneficial so more researchers who are interested in using the Emmental library.

I think Visual Domain Decathlon is a very common multi-task benchmark. Thanks

[senwu]

Thanks! I put this on my todo list.