I'm lucky enough to live close to one of the most beautiful inner-city parks in the world: Regent's Park in London. It is the perfect place to hang out any time of the year.

Especially in Summer though, many tourists flock to the place is full of tourists of all ages taking pictures of the beautiful John Nash designed vistas.

That's how I had the idea of buiding this model

Luckily we have just finished the first version of FieldDay, an app that allows anyone to collect data and train vision AI models from scratch in a matter of minutes.

Originally designed to scratch our own itch for data collection (mention internal tool angle) and rapid model iteration it's been embraced by Snap creators to build ML models for their lenses for marketing.

It's also a great tool to prototype real-time vision AI powered interactions in no-time, so that's exactly what I started to do.

I collected a few examples of just pure background and then a few with my finger in various positions around the frame.

One thing I noticed is that people might still want to take photos of their hands up-close, so I had to collect a few examples so that the model could differentiate between a true accident and just a hand that was intentionally captured.

Designing the interaction — making it feel Apple native and integration with the shutter button.

Talk about robustness and testing with a few more people in different places etc.

But you can try it for yourself — App Clip

Available on HuggingFace, GitHub