Places365 and Fashionpedia

Scene tags and garment detection for curator, each converted to ONNX by a curator command.

Used by

Package	What for	Reads
argus-curator (opens in new tab)	Scene tags (`places365-convert --arch resnet18\|resnet50`).	`ARGUS_CURATOR_PLACES365_MODEL`, `places365-convert`, `[places365-convert]`
argus-curator (opens in new tab)	Garments (`garments-convert`, from `valentinafevu/yolos-fashionpedia`).	`ARGUS_CURATOR_GARMENTS_MODEL`, `garments-convert`, `[garments-convert]`

Official guides

Places365 on GitHub (opens in new tab)
yolos-fashionpedia (model card) (opens in new tab)
Fashionpedia (opens in new tab)

Minimal setup

Each model is converted to ONNX once, into curator's model directory:

pip install 'argus-curator[places365-convert,garments-convert]'
argus-curator places365-convert --arch resnet18   # or resnet50
argus-curator garments-convert

To keep the files elsewhere, set ARGUS_CURATOR_PLACES365_MODEL and ARGUS_CURATOR_GARMENTS_MODEL to their paths. The converter extras are needed only for the conversion, not at scan time.