Digitizing human tactile perception
Can you feel it?
The smoothness of the cloth. The scratchy ridges of the tree. Isn't it incredible that you can feel the materials in the image above just by looking at them, without actually touching them?
Your brain is very good at processing the characteristics of visual cues and transforming these signals into understandings and feelings using the connective wiring of your brain. This processing, developed through brain evolution and life experience, allows you to really feel the tactile qualities of an object, although you are not physically doing so.
Imagine a brain-driven technology that could capture the human perception of how objects feel to the touch!
Brainvivo is on a journey of digitizing human perceptions, building a groundbreaking Humanized AI system that replicates how the human brain processes data such as the perception of tactile feel. Based on over 20 years of neuroscience research, Brainvivo’s cutting-edge technology unlocks the sophisticated biological network of the brain and generates predictive brain responses to automate tasks that require human decision-making, emotional insight, judgment, and intuition.
Brainvivo's 'Digitized Brain' simulates the human brain's response to data. To capture the cognitive process of tactile perception, Brainvivo trained the Digitized Brain on images of two fabrics known to have opposing tactile properties: silk and wool. Silk is smooth, and wool is rough. The model was trained using only 30 images per category and tested with 21 images per category, all with images of various colors, angles, shapes, and contexts. The Digitized Brain was able to distinguish between silk and wool with over 90% accuracy!
The human brain is a highly sophisticated network of interwoven subnetworks. The chart above shows the brain response simulated by the Digitized Brain, with each column representing the activity of one of 1000 simulated brain regions in response to images of silk and wool.
You don't need to be a neuroscientist to see the similarities and differences in the brain region activity in response to the two fabrics. Columns with similar activity form a network of brain regions capable of recognizing the similarities between silk and wool: both are fabrics with similar uses, such as for clothing. However, the columns with differing activity reveal another network that recognizes the differences between them - most importantly, the difference in the perception of their tactile properties.
Brainvivo validated that the Digitized Brain effectively captured tactile perception: without training it on any data beyond the images of silk and wool, this model was tested on pairs of completely new images of objects that are not fabrics (as seen below) and correctly distinguished between them with 85% accuracy!
The results of the Digitized Brain demonstrate that the model did not just learn to separate between the visual properties of silk and wool; it actually digitized tactile perception.
Using only images of silk and wool, the Digitized Brain learnt how to determine whether any object is rough or smooth!
The implications of this achievement is this: many other qualitative aspects of human perception can be automated and incorporated into a business solution, and Brainvivo can create a Digitized Brain that analyzes visual content through the scope of that perception!
Technological Advantages of the Digitized Brain
The Digitized Brain doesn't just distinguish between the physical features of silk and wool - it understands and contextualizes their properties to learn the difference in the perception of how they would actually feel!
Brainvivo was able to effectively digitize human tactile perception with relatively low CPU consumption and using only ~5% of the amount of data needed for classical machine learning models!
The Digitized Brain is highly adaptable - though it trained on specialised data, it was able to generalise and transfer what it learnt about silk and wool to distinguishing between 'smooth' and 'rough' objects in new images!
Brainvivo’s approach to building models is totally unique: focusing on the cognitive phenomenon of tactile perception, training the Digitized Brain on a relatively small amount of specialised data, and taking advantage of the model's innate capacity for transfer learning to produce a generalized model of that phenomenon. If this can be done for tactile perception, imagine all the different cognitive abilities that can be captured in the same way!