deeplabv3-mobilenet-int8-ov

OpenVINO IR export of torchvision/deeplabv3_mobilenet_v3_large, quantized to INT8 — 11 MB.

Quantization	weight compression (NNCF)
Lesson	02 Hello semantic segmentation
Built by	convert/convert_all.py of the ARCademy OpenVINO courseware

Output [1,21,H,W] (VOC classes); spatial dims are dynamic.

from huggingface_hub import snapshot_download
model_dir = snapshot_download("circulus/deeplabv3-mobilenet-int8-ov")