2 Comments
User's avatar
Daniel Popescu / ⧉ Pluralisk's avatar

Couldn't agree more. This genuinely marks a significant leap. The struggle with structured image analisys has been a real bottleneck. What if Moondream 3's capacity for structured outputs could streamline large-scale ecological surveys, counting specific species or tracking changes in biomass over time? The implications are huge.

Gilbert Mizrahi's avatar

Daniel, yes. This kind of applications with machine vision using generative AI, opens a lot of interesting applications. Many of them are possible today using traditional machine vision models, but they require a lot of preparation: labeling tons of images, and training.

Although there are services that make this simpler to the users, they still are time consuming, expensive and not something that anyone could do.