What’s Next#

ScubAI started simple — capture RAW, analyze the image, correct the colors. That already solves the core problem: shoot underwater, get the colors back. But the physics allows more, and we’re actively researching and developing several directions to push the correction further.

Better water models#

The current correction estimates water conditions from the image itself — color cast, channel ratios, attenuation patterns. This works, but the image is one data point. More information means better correction.

One direction: fetching real water conditions from third-party providers — ocean color data, chlorophyll concentration, turbidity measurements. If you’re diving at a known location, the app could pull current water characteristics and use them to inform the attenuation model before even looking at the image. The variation between water types is the single biggest source of uncertainty in any underwater correction — reducing that uncertainty at the input improves everything downstream.

Dive computer integration#

The app estimates the depth at which a photo was taken by reading light attenuation in the image. A Bluetooth-paired dive computer can provide that depth directly — matching photos to the dive profile timeline gives an exact value for each shot. Better input, better correction.

Calibration charts#

A different angle: bring a known reference into the scene. Photograph a calibration chart with known spectral reflectance at the start of a dive, and the app knows exactly how the water distorted each channel. Compare what the camera recorded with what the chart should look like, and the correction becomes measured rather than estimated. How the known patches were distorted also reveals the water’s optical character — blue water, green water, or something in between.

A dedicated article on calibration workflows and the chart itself is in the works.

ScubAI is in active development. Follow progress at scub.ai.

References#