MIT has published a news release about how neuroscientists in the McGovern Institute for Brain Research have recently made significant advances in their attempts to learn how the inferotemporal (IT) cortex identifies and categorizes visual data. The ability to visually recognize objects, while usually taken for granted because it happens quickly, automatically, and subconsciously, is actually a complex problem for the brain to solve. This research provides some insight into how the brain encodes, formats and saves visual information.
Captured first by the retina, visual information passes through several neural layers before reaching the highest visual level, the IT cortex. The IT cortex identifies and categorizes the visual pattern before sending it to other regions of the brain. The whole process takes a fraction of a second.
For the study, performed by James DiCarlo’s and Tomaso Poggio’s labs, researchers showed monkeys pictures of objects that could be easily categorized into groups, such as faces, toys or vehicles. The objects themselves varied in their size and position in the image. The scientists were able to record the IT neural output while the monkeys viewed the objects and were then able to build a huge database of neural firing patterns from the data. Using a sophisticated computer algorithm the scientists could correlate each object with a specific pattern. In addition to decoding the neural patterns they learned that only a split second’s worth of data was required to identify and classify the object regardless of whether the monkey had already seen that object at that size or position in the image.
The research was funded by DARPA, the Office of Naval Research and the National Institutes of Health and will appear in the November 4th issue of Science.
Read the full news release at the MIT News Office: “Neuroscientists break code on sight“