"NeuralLoom" is an interdisciplinary artwork that visualizes the intricate processes by which deep learning models perceive and interpret visual information. By showcasing how a convolutional neural network processes image data, the project offers insights into the journey of image features through the model's layers, bridging the gap between machine computation and human understanding.
The artist captures intermediate outputs from each activation layer of a deep learning model during the classification of diverse images representing various categories. This vast data, organized into multi-dimensional tensors, reflects the patterns generated by the model when processing different visual inputs. The variations in these patterns became a focal point for visualization.
Presented as an interactive 3D installation, viewers can navigate through the neural network's layers and channels. Interactive elements allow users to toggle between layers, display underlying data values, and adjust visualization parameters. As they explore, viewers observe emergent patterns that signify how the model recognizes and interprets various aspects of the images, illustrating the progression from simple features to complex structures.
"NeuralLoom" makes these abstract processes tangible, fostering a deeper appreciation of the technologies that influence our perception. It invites reflection on the evolving relationship between technology and humanity, encouraging dialogue about the role of artificial intelligence in society and its impact on human cognition and creativity.
.png)
.png)
