Vlads stealth - LinkedIn post17 Aug 2025 13:51
From Rybak’s Neural Preprocessor to the Next Frontier of Visual AI
I’m excited to share a bit of my journey and why I believe we’re on the cusp of a profound shift in visual AI.
During my undergrad studies, I, under the guidance of Prof. Ilya Rybak, co-developed a Neural Network Visual Preprocessor, a parallel, visual cortex-inspired system that builds on Ilya Rybak’s (https://lnkd.in/ecr4H3hw) neural model, aimed at accelerating and enhancing low-level visual processing. My work adapted the orientation selective visual cortex-inspired architecture, where anisotropy of inhibitory feedback from interneurons and multi-layer contour segmentation serve as the computational bedrock for invariant image recognition https://lnkd.in/eMMxWmU5
This work is related to the Pulse‑Coupled Neural Network (PCNN) framework: a biologically inspired, two-dimensional spiking neural model proposed by Eckhorn et al. (1989) and formalized via the Rybak et al. models from the early 1990s https://lnkd.in/eTswEr_p
PCNNs produce time-based pulse outputs via iterative dynamical signal processing and offer innate robustness in handling noise, geometric variations, and subtle intensity shifts—making them a high-performance tool for image segmentation, feature detection, medical image fusion, and more.
PCNN variants—such as dual‑channel Rybak networks—have already shown impressive results in medical imaging tasks, including MRI/CT fusion.
https://lnkd.in/eymSrYuf
More broadly, recent reviews highlight PCNN’s success across a wide array of domains, from edge detection and image compression to remote sensing and object recognition https://lnkd.in/eA3PfWGA
Why This Matters Now
I’m thrilled to announce that our stealth-stage company is building upon this rich foundation—melding the Rybak/PCNN principles with modern computational architecture to deliver practical, real-world AI image processing solutions that push us closer to general artificial intelligence.
We leverage PCNN-inspired temporal dynamics and spiking mechanisms for resilient and adaptive preprocessing.
Our approach embraces bio-inspired design for energy efficiency, speed, robustness, and interpretability.
The long-term vision? Positioning PCNN-centric computation as a cornerstone for generalist visual AI systems, systems capable of perceiving, learning, generalizing, and reasoning across diverse visual contexts.
What’s Next
We’re gearing up for building an MVP for applications in medical imaging, anomaly detection, and beyond.
We’re open to speaking with collaborators and forward-thinking investors who recognize the power of biologically inspired neural architectures.
If you’re passionate about cutting-edge AI, neuroscience-inspired models, or reshaping how machines perceive and learns about the real world—let’s connect.
Sign In
Follow the stocks
