Assessing Computer Vision Models with Roboflow's Piotr Skalski

EPISODE SUMMARY

Piotr discusses his criteria for evaluating computer vision models, as well as a breakdown of what makes Meta's recent "Segment Anything" Model exciting.

EPISODE NOTES

Today’s guest is a Developer Advocate and Machine Learning Growth Engineer at Roboflow who has the pleasure of providing Roboflow users with all the information they need to use computer vision products optimally. In this episode, Piotr shares an overview of his educational and career trajectory to date; from starting out as a civil engineering graduate to founding an open source project that was way ahead of its time to breaking the million reader milestone on Medium.We also discuss Meta’s Segment Anything Model, the value of packaged models over non-packaged ones, and how computer vision models are becoming more accessible.Key Points From This Episode:

• What Piotr’s current roles at Roboflow entail.
• An overview of Piotr’s educational and career journey to date.
• The Medium milestone that Piotr recently achieved.
• The motivation behind Piotr’s open source project, Make Sense (and the impact it has had).
• Piotr’s approach to assessing computer vision models.
• The issue of lack of support in the computer vision space.
• Why Piotr is an advocate of packaged models.
• What makes Meta’s Segment Anything Model so novel and exciting.
• An example that highlights how computer vision models are becoming more accessible.
• Piotr’s thoughts about the future potential of ChatGPT.

Tweetables:“Not only I showcase models but I also show people how to use them to solve some frequent problems.” — Piotr Skalski “I am always a fan of models that are packaged.” — Piotr Skalski “We are drifting towards a direction where users of those models will not necessarily have to be very good at computer vision to use them and create complicated things.” — Piotr Skalski