Neural architeuct generation using Diffusion modeling
In this project we will exemine changing NN architecture to be more robust / less computationaly heavy, using classifier free diffusion models
Posts by Collection
projects
Learn with Noisy Labels in segmentation
Investigate the biasses of self annotation and how to combat noisy labels in segmentation tasks
Real-time Instance Segmentation
Formulate the segmentation task as regression only
Mixing augmentation for semi supervised learning
Generate new insights about how and what to mix
Increase the latency of LLM
By adverserial attack
Semi Supervised Segmentation for 3D cardiac data
Levreging previus work
publications
Bimodal-Distributed Binarized Neural Networks
Published in Mathematics 10(21), 4107, 2022
Shape weight distributions to preserve useful representations under binarization.
Download here
AMED: Automatic Mixed-Precision Quantization for Edge Devices
Published in Mathematics 12(12), 1810, 2024
Choose mixed precision with the target hardware in mind.
Download here
Semi-Supervised Semantic Segmentation via Marginal Contextual Information
Published in TMLR, 2024
Use neighboring pixels together to refine pseudo labels and learn more from unlabeled images.
Download here
Hysteresis Activation Function for Efficient Inference
Published in NeurIPS ENLSP Workshop · PMLR 262, 2024
Revisit the activation function to address dying ReLUs with minimal inference complexity.
Download here
Robot Instance Segmentation with Few Annotations for Grasping
Published in WACV, 2025
Learn from scene changes and unlabeled images to reduce annotation demands for robotic grasping.
Download here
Maximal Brain Damage Without Data or Optimization: Disrupting Neural Networks via Sign-Bit Flips
Published in TMLR, 2026
A handful of critical sign bits can break a neural network. Finding them reveals where protection matters most.
Download here
talks
Talk about efficient models
Published:
I gave a talk to the workers of Linvo about efficient modeling, compration, quantization and NAS, challenges and current state-of-the-art on edge devices.
teaching
Computer vision
Workshop, Technion, EE faculty, 2021
Teaching 046746 - From SIFT and templete matching to Deep learning for tracking
Deep Learning
combined Undergrad and Grad course, Technion, CS + Raichman, CS, 2022
Head TA for cs236781. I am the head TA, means all tutorials and home assignments under my responsibilities. We’ve created a course with three segments