Quality Control
The Challenge: While quality control (QC) is a vital part of the production workflow, it requires a lot of precise, time-consuming work that is heavily reliant on manual checks and balances to ensure consistency.The Solution: An AI co-pilot that works with retouchers and other studio professionals to help exercise faster, cleaner, and more standardized quality control across all produced content.
Our focus is on ensuring our customers deliver the best possible work, as effectively as possible, and at the highest standard. So, developing AI to assist in quality control is high on our list. We make an effort to pre-flag any concerns so the person responsible for QC gets a really efficient co-pilot to catch any issues.
There are a lot of places to help with QC, and among the ones we are building models to assist with are:
Focus
Are the right things sharp and in focus? We want those eyes, details, or whatever is most important for your shot to be sharp as a tack. Using AI to identify the important parts of the image and image analysis to determine sharpness, we can determine if focus is correct for the image and flag any signs that it is not.
Color validation
Colors can change, both on set and in post-production. But have no fear—our AI will catch discrepancies between images in the same series and products compared to the defined hex color or sample.
Crops and composition
You want to be consistent with your cropping and composition. So why not let an AI look at that before you spend time time validating manually.
Grain and noise
Clean images look…. well, cleaner. So we are checking for grain, noise, and compression artifacts.
Auto markup
Dust, dirt, flyaways, reflections, blemishes, and all the other usual suspects… we will help by marking up where the AI sees any of those things and letting you send those comments off to post-production for fixing.
As you can see, we are on a continuing journey to help our users catch any detail in the QC process while also speeding up the workflow tremendously. To make that happen, we use different vision processing systems, deep pixel analysis, and many other tricks of the AI trade.