Adriano Cardace

Coming to #ICLR2024: what's the best way to extract information from implicit neural representations (INRs)? We show that tri-planes alone are enough to effectively classify or segment 3D objects in neural fields. 🧵

Can we process hybrid neural fields to solve downstream tasks such as 3D object classification/segmentation? In this new work, we show that using hybrid representations such as tri-planes can provide many advantages over neural fields parametrized as MLP: arxiv.org/abs/2310.01140

Signals are networks so networks are data and so networks can process other networks to understand and generate signals. Puzzled ? Check out out our #iclr2023 paper (cvlab-unibo.github.io/inr2vec/) Deep Learning on Implicit Neural Representations of Shapes

Happy to release the code for our new paper "Self-Distillation for Unsupervised 3D Domain Adaptation" accepted at WACV2023! Code and paper can be found here: cvlab-unibo.github.io/FeatureDistill… A big thanks to my colleagues @pierluigi92 and @RSpezialetti

“Yeah, but they haven’t demonstrated they can go *up* the stairs.” x.com/slender_sherbe…

La situazione sta un pelino sfuggendo di mano #Chiellini #Saka