Show HN: I turned my face rec system into a video codec (vertigo.ai)

Article summary

The article discusses a face recognition system that has been turned into a video codec, allowing for efficient video compression. The system is designed to work on edge devices and can handle low-bandwidth connections. The developer claims that the system can capture detailed facial features, including scars and facial expressions. The system's performance and potential applications are discussed in the comments.

Main themes

• face recognition
• video compression
• AI bias
• patent law
• video conferencing
• surveillance and tracking

What commenters say

• The face recognition system is impressive and has potential applications in video conferencing and other fields.
• The system may have limitations and biases, particularly with regards to diverse skin tones and facial features.
• Including a diverse range of faces in the training data is crucial to ensure the system works accurately for all users.
• The system's ability to handle non-facial objects and scenes is also important for its practical application.
• The developer's approach to patenting the invention is questioned, with some arguing that mathematical formulas cannot be patented.
• The system's potential to address issues like unwanted explicit content in video chats is also discussed.
• Some commenters are concerned about the potential misuse of the technology, particularly in surveillance and tracking applications.