RT @ShambhaviCodes: (5/5) Checkout the paper : GitHub : Thanks to the co-authors : @93Balam….

RT @josedolz_ets: It seems some of my students will go to Hawaii this winter to present their works at @wacv_official .1) Prompting classes….

RT @RoozbehMottaghi: If you are a recent PhD graduate or will graduate soon and have a passion for Embodied AI and Computer Vision, join ou….

RT @Sanjusinha7: I am hiring multiple Computational Biologists for my new lab @sbpdiscovery , San Diego. An alternative to a postdoc, it w….

Paper: Code: Authors: @93Balamuralim, @sukeshadiga, @bing_bingyuan, Herve Lombaert, @IsmailBenAyed1, @josedolz_ets.

* Based on these observations, we propose a principled and simple solution based on equality constraints on the logit values, which enables to control explicitly both the enforced constraint and the weight of the penalty, offering more flexibility.

* Our analysis shows that SVLS lacks a mechanism to balance the contribution of the constraint with the primary objective, potentially hindering the optimization process.

* We present a constrained optimization perspective of SVLS and demonstrate that it enforces an implicit constraint on soft class proportions of surrounding pixels.

* Ensuring reliable confidence scores from deep networks is important in the medical domain. * Spatially Varying Label Smoothing (SVLS) models ucertainty by softening the pixel label assignments with a discrete spatial Gaussian kernel.

Happy to inform that, our paper "Trust your neighbours: Penalty-based constraints for model calibration" has been accepted in MICCAI 2023. The reviewers provided encouraging comments and were impressed with our novel formulation and solution. #MICCAI2023.

RT @ipmi2023: Second oral talk of Day 5 beginning now, on harmonization flows.

RT @josedolz_ets: A lot of interest in our second poster for calibrating deep neural networks #CVPR23 poster 354 @IsmailBenAyed1 @jerome_ro….

RT @josedolz_ets: If you are at #CVPR2023 drop by to see our posters for tomorrow. AM: 289 - A strong baseline for generalized few-shot se….

RT @josedolz_ets: One of our latest works on calibration in segmentation networks has been accepted at MedIA. Kudos to all the team @93Bala….

Awesome series.

RT @midl_conference: The #MIDL2020 proceedings are now online! You can find them on the @JmlrOrg website:.

RT @esindurmusnlp: It’s great that schools are waiving the GRE requirement. But I remember the main cost that prevented me from applying to….

RT @dennybritz: A Wholistic View of Continual Learning: If you're interested in Continual Learning and Active Lear….

RT @wanmeher: New researcher, trying to publish in a high impact journal. @OdedRechavi

RT @zhoubolei: Our recent work on disentangling the latent variation factors in generative models in unsupervised manner. Fun to interact w….