LibreUNets-sem

U-Net (UNet-S5-D16 with an FCN head) Cityscapes 19-class semantic segmentation weights, converted for LibreYOLO.

Source

Derived from open-mmlab/mmsegmentation at commit b040e147adfa. The source checkpoint is the official fcn_unet_s5-d16_4x4_512x1024_160k_cityscapes_20211210_145204-6860854e.pth (SHA-256 6860854ebe657b0f85f6ec0bf4315fca3b54e6ce639710e76ab055ebc292c090). Copyright 2020 The MMSegmentation Authors. Licensed under the Apache License 2.0.

Modifications

Converted with weights/convert_unet_weights.py in the LibreYOLO source repository. The state-dict layout is upstream's and the learned parameters are unchanged; the converter wraps the tensors with LibreYOLO checkpoint metadata (model_family=unet, size=s, task=semantic, nc=19, evaluation canvas 1024x2048).

Usage

from libreyolo import LibreYOLO

model = LibreYOLO("LibreUNets-sem.pt")
result = model("street.jpg")

License

Apache License 2.0. See the LICENSE and NOTICE files in this repository.

Dataset Note

Cityscapes has separate dataset terms and is not redistributed in this repository.