Kellify’s scientific approach builds upon the similarities between human perception and the behavior of artificial neural networks. Just like neurochemical processes taking place within the human nervous system transform information sensed by our retinas into shapes, colors, patterns, and the entire complexity of visual forms, Deep Learning algorithms capture and extract critical features of a visual object and, through its further elaboration, provide a wholesome understanding of the sophistication of an image. Through this data-driven approach, we are able not only to identify the most captivating image, but also to pinpoint the most distinctive and compelling areas within each frame.
Visual stimuli play a significant role in human cognition. Images – being our primary thoughts modality – provide a vivid and detailed representation of the surrounding environment. Visual information is easily accessible and comprehensible to everyone due to its universal and intuitive character, and the capacity to create a strong emotional impact. Affective characteristics incorporated within a visual frame provide the observer with immediate information about the meaning and importance of the perceived image. Early integration of sensory and affective information about a visual stimulus allows the viewer to assign value and determine further engagement or dismissal of the object of perception.
The activation of human senses is a result of an interplay between various sensory and affective characteristics of the object of perception. While its emotional value is being determined mostly by the semantic content, the initial ultra-rapid (~ 50 msec) phase of perception includes the analysis of low-level features, such as, color, composition, contrasts etc. These intrinsic properties of an image stimulate the recruitment of human attentional circuits. Attention is indispensable for the performance of further perceptual tasks, judgment formation and decision-making processes. Making predictions about the impact of sensory characteristics of an image on human perception tends to be elusive. For this reason, at Kellify we came up with data-driven, automated methods enabling us to understand which images stimulate human senses, evoking engagement of the viewer.
Kellify believes in the importance of addressing the issue of algorithmic bias and accountability. We put it into practice by working on large-scale datasets, ensuring meaningful aggregate analyses. Being concerned about the representativeness of the data, we step away from individual, one-to-one user observation and instead dive into high volumes of data coming from a variety of sources. We are invested in providing highly reliable, powerful solutions, through combining algorithmic decision making with human expertise.