News
[2025/01/24]
We released our code (v1.0) on GitHub!
Here are two key features: (details covered in this tutorial)
Here are two key features: (details covered in this tutorial)
- - 7 minutes for the sandbox cortex model training using a single RTX 4090 GPU.
- - 2 hours for the main cortex model training using a single RTX 4090 GPU.
Subscribe
Sign up here for updates and more! We'd appreciate any feedback you'd like to share!
A Computational Framework for Modeling
Emergence of Color Vision in the Human Brain
ICLR 2025 (Oral)
Atsunobu Kotani & Ren Ng
Department of Electrical Engineering and Computer Sciences,
University of California, Berkeley
Two Sentence Summary
This work addresses the long-standing mystery of how the brain constructs color vision with correct color dimensionality, purely from optic nerve signals.
By simulating the eye to generate biologically realistic optic nerve signals from hyperspectral images and integrating a bioplausible cortical learning mechanism, we demonstrate that color vision can naturally emerge, solely from these signals.
Supplementary Video
This supplementary video offers a concise visual overview of our methodology and key findings, and will likely enhance your understanding of the paper. We highly recommend watching it!
Tutorials
To reproduce the results presented in this work, please visit our Tutorials (click the button below↓)