Multi-task Learning with 3D-Aware Regularization

We propose a novel regularization method that can be plugged into diverse prior Multi-task Learning architectures for dense vision problems including both convolutional and transformer networks where the structured 3D-aware regularizer interfaces multiple tasks through the projection of features extracted from an image encoder to a shared 3D feature space and decodes them into their task output space through differentiable rendering.

Multi-task Learning with 3D-Aware Regularization,
Wei-Hong Li, Steven McDonagh, Ales Leonardis, Hakan Bilen,
ICLR 2024 (arXiv 2310.00986)

Updates

• January'24, Our paper is accepted to ICLR'24! Code will be available soon!

Contact

For any question, you can contact Wei-Hong Li.

Citation

If you use this code, please cite our papers:

@inproceedings{li20243dawaremtl,
    author    = {Li, Wei-Hong and McDonagh, Steven and Leonardis, Ales and Bilen, Hakan},
    title     = {Multi-task Learning with 3D-Aware Regularization},
    booktitle = {International Conference on Learning Representations (ICLR)},
    month     = {May},
    year      = {2024}
}