Update README.md

README.md

@@ -1,3 +1,77 @@
----
-license: cc-by-4.0
----
+---
+license: cc-by-4.0
+datasets:
+- danielritchie/cinematic-mood-palette
+language:
+- en
+tags:
+- tflite
+- embedded
+- emotion
+- color
+- hri
+- robotics
+- affective-computing
+- real-time
+- vad
+- tiny-model
+---
+
+# VIBE Color Model
+
+A 365-parameter TFLite model that maps emotional state to cinematic color expression. Designed to run on embedded hardware with minimal compute.
+
+## Model Description
+
+Given a 5-dimensional emotional coordinate (VAD+CC), returns a cinematic visual treatment — not just a color, but RGB plus independent Energy and Intensity parameters drawn from cinematographic practice.
+
+**Architecture:** 5→16→12→5 fully connected network
+**Size:** 3.5KB
+**Parameters:** 365
+**Format:** TFLite (embedded deployment), H5 (inspection/fine-tuning)
+
+## Inputs and Outputs
+
+**Input:** VAD+CC vector — 5 float values in [0, 1]
+
+| Dimension | Meaning |
+|---|---|
+| Valence | Negative ↔ Positive emotional tone |
+| Arousal | Calm ↔ Energized |
+| Dominance | Passive ↔ Powerful |
+| Complexity | Minimal ↔ Rich |
+| Coherence | Chaotic ↔ Harmonious |
+
+**Output:** 5 cinematic parameters — 5 float values in [0, 1]
+
+| Dimension | Meaning |
+|---|---|
+| R | Red channel |
+| G | Green channel |
+| B | Blue channel |
+| Energy | How alive/active the display feels |
+| Intensity | How pronounced the effect is applied |
+
+## Training Data
+
+Trained on [danielritchie/cinematic-mood-palette](https://huggingface.co/datasets/danielritchie/cinematic-mood-palette) — ~80 curated anchor points mapping emotional states to visual treatments drawn from film and photography.
+
+## Validation
+
+Validation is qualitative. The model is evaluated by behavioral coherence — does the output feel cinematically appropriate for the emotional input? Formal quantitative benchmarks are not meaningful for a model of this size and purpose.
+
+## Intended Use
+
+Part of [VIBE-Eyes](https://github.com/brainwavecollective/vibe-eyes) — a real-time emotional display system for conversational robots. The model runs on-device, receiving VAD+CC vectors from an edge emotion engine and driving LED color output without any cloud dependency.
+
+Also useful as a lightweight reference implementation for anyone mapping affective state to visual expression in constrained environments.
+
+## Limitations
+
+- Small training set (~80 anchor points): functions as a reference structure, not comprehensive coverage
+- Culturally specific: draws primarily from Western cinematic tradition
+- Interpretive: mappings reflect observed patterns in film, not objective measurements
+
+## License
+
+CC-BY-4.0 — use freely with credit