[Models] - Aggregated issue #1564

Tracks model additions. Backlog items become sub-issues when someone picks them up.

Backlog

Audio

• Sound event classification (YAMNet, CED-tiny, CLAP)
• Pitch / note transcription (CREPE-tiny, basic-pitch)

NLP

• Zero-shot NER (GLiNER)
• Zero-shot / sentiment text classification (DeBERTa-v3 NLI, ModernBERT)
• Punctuation and truecasing for STT output

Vision

• Super-resolution / low-light enhancement (Real-ESRGAN compact, Zero-DCE++)
• Hand landmarks / gestures (MediaPipe Hands)
• Face recognition embeddings (MobileFaceNet)
• Document edge detection and dewarping (DocAligner, DocTr)
• Handwriting recognition (TrOCR)
• Video action recognition (MoViNet-A0 stream)

Sensors

• Human activity recognition from IMU (TinyHAR)