The DawaImg blog

Notes on offline face forensics.

Practical, privacy-first writing on the craft: chain of custody, scale, clustering, redaction, and doing this work responsibly on your own machine.

The phrase pairs face analysis with data control. Here’s what each means in practice.

Why “where it runs” is the most important architectural decision you’ll make.

How a SourceRecord → FrameRecord → FaceRecord model turns “faces” into defensible records.

What parallelism, resumability, and disk-backed caches actually buy you at document scale.

Face clustering should group appearances, not claim identities. Here’s how to think about it.

Blurring, redaction, and controlled exports — protecting subjects without losing the analysis.

Understanding how face systems misfire is just as useful as knowing how they find faces.

Real cases are USB sticks, SSDs, and network shares. How to structure messy media.

The single most impactful reviewer setting. How to pick it, and how to sanity-check it.

A software’s constraints shape what it can be misused for. On closing the code and doing this ethically.

Read the ideas. Then try the tool.

Every article points back to a workflow you can run on your own machine today.