FieldDay is the pick when the real problem isn’t finding an existing scanner, but building one that understands your specific world. Unlike Brickit’s fixed LEGO-focused recognition, FieldDay lets you capture examples, iterate, and train a custom vision model for the exact classes you care about.
That custom approach is useful when off-the-shelf recognition breaks down—similar-looking items, unusual parts, or niche categories. Instead of hoping a general model gets better, you can systematically improve performance by collecting better training data and refining the model as your needs evolve.
FieldDay also fits nicely into automation-driven workflows: once your model is reliable, you can trigger actions via phone-centric integrations (like Siri Shortcuts or connected workflows), making it feel more like a personal “vision sensor” than a single-purpose app. One practical consideration is that training currently runs
on cloud GPU, for now, which can matter if you’re aiming for fully offline iteration.
If Brickit is about immediate build suggestions, FieldDay is about owning the recognition problem end-to-end for whatever you’re trying to identify.