our demo team is getting ready #ECCV2024

I'll be at #ECCV2024 this year, where Mariam Hassan will be presenting her work on Thermal+RGB imaging using NeRF (https://malcolmmielle.github.io/publication/thermonerf) at the workshop on Neural Fields Beyond Conventional Cameras (https://neural-fields-beyond-cams.github.io/). This work is a collaboration with Florent Forest and Olga Fink of the IMOS Lab - EPFL.

Find the code on Github (https://github.com/Schindler-EPFL-Lab/thermo-nerf), the dataset on Zenodo (https://zenodo.org/records/10835108), and the pdf on arxiv (https://arxiv.org/pdf/2403.12154)

Come talk to us!

🚀 Thrilled to Share Some Great News from Our Research Team!!1! 🚀

Say hello to balanced datasets and enhanced model performance!

We're excited to announce that our paper, “Dynamic Label Injection for Imbalanced Industrial Defect Segmentation,” has been accepted to the VISION workshop at #ECCV2024, hosted by #apple ! 🎉

Our research presents DLI (Dynamic Label Injection), a novel method designed to address the challenge of imbalanced multi-class semantic segmentation in deep learning systems. DLI works by rebalancing class distribution in training batches, utilizing Poisson-based seamless image cloning and cut-paste techniques to transfer defects effectively. Our experiments on the Magnetic Tiles dataset demonstrate that DLI outperforms other balancing loss approaches, even in weakly-supervised settings.

A big shoutout to my coauthors Emanuele Caruso from Free University of Bozen-Bolzano and my colleague Alessandro Simoni, Ph.D for their exceptional contributions! 🙌 <3

Check out our preprint here: https://arxiv.org/abs/2408.10031 📄
Explore our code on GitHub: https://github.com/covisionlab/dynamic-label-injection 💻

#deeplearning #artificialintelligence #industry

7th Workshop on Computer Vision for ART Analysis at ECCV 2024 in Milan! Explore the intersection of #ComputerVision and art, from generative techniques to 3D reconstructions of historic sites. Submit your papers by July 3!
https://cmt3.research.microsoft.com/VISARTVII2024
#ECCV2024 #VISART