STUDIOLAB ML inference Package

Install

• pip install studiolab-ml

RUN

All input image type is PIL Image

MLFT

from studiolab_ml import MLFT

mlft = MLFT()
out = mlft.predict(img, cat_id)

• result is same dict type as "get_attributes" in ML-API

Pose Compo

from studiolab_ml import PoseCompo

pcp = PoseCompo()
out = pcp.predict(img)

• output examples
• outfit image - {'cut': 'outfit', 'background': 'blind', 'direction': 'front', 'head': 'head', 'part': 'full', 'pose': 'stand', 'detail': None}
• product image - {'cut': 'product', 'background': None, 'direction': 'front', 'head': None, 'part': None, 'pose': None, 'detail': None}
• detail image - {'cut': 'detail', 'background': None, 'direction': None, 'head': None, 'part': None, 'pose': None, 'detail': [shoulder, sleeve, ..]}
• noise image - {'cut': 'noise', 'background': None, 'direction': None, 'head': None, 'part': None, 'pose': None, 'detail': None}

FIC

from studiolab_ml import PoseCompo

infer = FIC(api_key)
res = infer(attribute_dict, user_inputs_dict)

• input and result is same dict type as "get_gpt_content" in ML-API

TODO

• create model cloud storage
• model download from cloud
• GPU inference