← Catalog

Digital twin · proof of concept

Mapping images to emotions

Capture face signals with ml5, store as JSON, pair each mood with a straight-ahead PNG, crossfade on keys 1–5. First step toward people in landscapes who react to a visitor’s camera.

Markdown · JSON · HTML · JavaScript · ml5 · Cursor

Live — POC

• Portrait replay Primary — mood crossfade · five PNGs · keys 1–5
• Face signal study Capture — webcam + record JSON + save still (S)
• Wire replay Skeleton — JSON drives wire head (wrap images later)
• Thought replay Mind layer — mood crossfade + thought-map.json · keys 1–7
• Witness Mirror (5s) Testimony + machine witness · DeepFace 5s samples · evaluation
• Witness Mirror (1s) Moment audit — DeepFace every 1 second · masking, grief vs fear
• Witness Mirror (live) Webcam + face-api · point vs counterpoint · browser lab
• Witness Mirror (DeepFace live) Trusted witness · local Python server · 15s story

Next

• Landscape test Place inhabitant in walkable terrain — billboard PNG or video
• Walking video Loop of you walking — body layer (`sessions/walk-loop.mp4`)
• Sidewalk video Third-person street view — witness clip (`sessions/sidewalk.mp4`)

Future

• Skeleton + wrap ml5 wire frame → bind mood images (skinning) → visitor camera drives reaction

emotion-map.json · catalog README