Related Experiment Video
Updated: Aug 5, 2026
07:08
Investigating Object Representations in the Macaque Dorsal Visual Stream Using Single-unit Recordings
Published on: August 1, 2018
STSBench: A Large-Scale Dataset for Modeling Neuronal Activity in the Dorsal Stream of Primate Visual Cortex
Ethan B Trepka1, Ruobing Xia2, Shude Zhu2,3
1Neuroscience Interdepartmental Program, Stanford University, Stanford, CA.
Arxiv
|July 29, 2026
Summary
Researchers developed STSBench, a large dataset of neural recordings from the primate dorsal stream. This resource aids in understanding visual processing and building better computational models for spatial and motion perception.
Area of Science:
- Neuroscience
- Computer Vision
- Computational Neuroscience
Background:
- The primate visual system has distinct ventral (object recognition) and dorsal (spatial/motion processing) streams.
- Convolutional Neural Networks (CNNs) excel at modeling the ventral stream but not the dorsal stream due to data limitations.
Purpose of the Study:
- To address the lack of large-scale datasets for the dorsal stream.
- To create a benchmark dataset for developing and evaluating computational models of dorsal stream function.
Main Methods:
- Collected large-scale, single neuron recordings from over 2,000 neurons in the superior temporal sulcus (STS) of Rhesus macaques.
- Recorded neural activity while macaques viewed thousands of unique, natural videos.
Main Results:
- Introduced STSBench, a dataset representing a nearly 50-fold increase over existing dorsal stream datasets.
- Demonstrated STSBench's utility for benchmarking encoding models of dorsal stream neuronal responses.
- Showcased the dataset's capability for reconstructing visual input from neural activity.
Conclusions:
- STSBench significantly advances the study of the dorsal visual stream.
- The dataset facilitates the development of more sophisticated models for spatial and motion perception.
- Enables deeper insights into the neural mechanisms of visual processing in the dorsal stream.
