Bill's approach to PoseNet was not one of wonderment at its capabilities, but of introspection. His query, “What do I want from it?” is a profound reminder that technology, irrespective of its sophistication, should be a conduit for artistic expression and not the primary spectacle. It’s an essential lesson in an age where the allure of new technologies can sometimes eclipse the very narratives they're meant to enhance. The technology, as advanced as it may be, played a supporting role, underscoring the narratives rather than overwhelming them.
The "Humans of AI" essay stresses the importance of understanding the context behind images. PoseNet, while sophisticated, still only detects poses without any understanding of the story or emotion behind the movement. For instance, the same pose could be interpreted differently depending on its context— a gesture could be seen as a dance move in one setting, a martial arts stance in another, or even a sign of distress in yet another. This lack of deep contextual understanding mirrors the essay's sentiment about the COCO dataset. Like the COCO dataset, PoseNet is trained on vast amounts of human images to accurately predict poses. Reflecting on the essay, I would wonder about the origin of these images, the consent of those pictured, and the diversity and inclusivity of the data used. Given the intimate nature of PoseNet's task—detecting and analyzing human bodies—these questions might become more pertinent to consider about.
Code Documentation
The original PoseNet application focused only on the interaction with the cat emoji and sound. My modifications to this code introduced an additional interaction with a dog emoji, playing respective sounds based on the user's proximity to each emoji. Additionally, visual feedback was added in the form of cat ears to enhance user experience. The essential part of the code calculates the distance between the user's eyes and uses this distance to proportionally draw cat ears above the eyes when the cat sound is playing. The ears adjust in size based on the user's distance from the camera, ensuring a consistent visual representation regardless of the user's position.
Comments