Dividing Hair into Real Sections

Me:

Claude, nice. The hair segmentation part is finally working. Great job.

But looking at the current results… when I make the hair longer, the entire visible area grows all at once.

Claude:

Right now, the editor is organized by four camera views: Front, Left, Right, and Back.

Instead, we should divide the hair into real salon sections.

For example:

Crown
Front
Temple
Side
Occipital
Nape

That’s the terminology professional hairstylists actually use.’

Me:

That actually makes much more sense.

If hairstylists already work this way, we should too.

Claude:

Exactly.

Hair isn’t cut based on where the camera is.

It’s cut section by section.

The rendering for each view can simply become a combination of those shared sections:

Front: Crown + Front + Temple

Left: Crown + Temple + Side + Occipital + Nape

Right: Crown + Temple + Side + Occipital + Nape

Back: Crown + Occipital + Nape

Strengthening the Connection Between Sections

Me:

Looking at this structure… I think we’ll eventually run into another problem.

If users don’t take the front, side, and back photos from exactly the right angles, the sections won’t line up properly.

Claude:

Exactly.

A “left-side” photo is rarely taken at a perfect 90° angle.

The front photo might be rotated a few degrees.

The back photo could be tilted slightly.

Even small differences like these cause the section boundaries to shift.

Instead of requiring perfect photos, we can estimate each camera angle using facial landmarks and automatically correct the pose before segmentation.

Me:

Oh… that’s clever.

Building a full 3D system would require much more computing power.

If we can solve the alignment problem first with automatic pose correction, that’s a much more practical approach.