Shinseungback Kimyonghun is a Seoul based artistic duo consisting of engineer Shin Seung Back and artist Kim Yong Hun. Their latest project Analytic Portrait, pictured above and featured (last) in the series of works that explore the representation (and computation) of human face through the use of computer algorithms (2012-2024+).
Cloudface, 2012
..is a compilation of cloud images that AI identifies as human faces, a phenomenon born out of AI’s error. While humans may also perceive faces in some of these images, they don’t regard them as actual faces. Instead, humans tend to imagine faces within the clouds. ‘Cloud Face’ explores the error and imagination of AI and humans.
On display at Ars Electronica Center, 2014-2021
Portrait, 2013
..is a series of digital portraits representing an identity (or a face) of a movie. Software detects faces from every 24 frames of a movie, and creates an average face of all found faces. These faces are layered to produce an “average” / centric figure(s) and the visual mood of the movie.
Cat or Human, 2013
Human faces recognized as cat faces by a cat face-detection algorithm.*
Cat faces recognized as human faces by a human face-detection algorithm.**
Computer vision library:
*KITTYDAR Face detection for cats
**OpenCV’s face detector
Nonfacial Mirror, 2013
The mirror avoids faces.
One can look at his/her face in the mirror only when it is a nonface.
Facial mirror, webcam, servomotor, Arduino, computer, face detection algorithm, custom software
Nonfacial Portrait, 2018
The artists invited painters and asked them to make portrait paintings of a person. The painters had to follow one rule: “the face of the portrait must not be detected by AI.”
While the portrait is being made, a camera equipped with three face detection algorithms* watches the canvas and a monitor informs the painter if any face is detected. The face of the completed portrait should not be visible to any of the algorithms.
It is tricky. If one paints the portrait close to the human subject, AI will easily detect the face. And the more one tries to make it unrecognizable by AI, the less the painting would be seen as a portrait of the person. One must find the small visual space that only humans can perceive.
In the installation, each portrait is presented along with the video of the painter’s painting process.
* OpenCV Face Detection, Dlib Face Detection and Joint Cascade Face Detection and Alignment.
Analytic Portrait, 2024
..is a series of portraits created based on the way AI perceives humans. 14 AI vision algorithms* are used to analyze the human subject, and the integrated results are recorded to construct the portrait image.
* Distinctive Image Features from Scale-Invariant Keypoints, OpenCV Face Detection, Dlib Face Detection, DSFD, Style Aggregated Network for Facial Landmark Detection, Keypoint Communities, Distribution-Aware Coordinate Representation for Human Pose Estimation, Ad-Corre: Adaptive Correlation-Based Loss for Facial Expression Recognition in the Wild, Instances as Queries, FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age for Bias Measurement and Mitigation., L2CS-Net: Fine-Grained Gaze Estimation in Unconstrained Environments, Real-time deep hair matting on mobile devices, Multi-ethnic MEBeauty dataset and facial attractiveness assessment, and Google Cloud Vision API
For more projects by Shinseungback Kimyonghun, see https://ssbkyh.com
