Manga109 speech-bubble segmentation (SafeTensors)
This repository contains a SafeTensors conversion of
huyvux3005/manga109-segmentation-bubble
for the Torch implementation in Koharu.
- Source revision:
f9a4108c4955136a810e5e92207972f3fb3a65fd - Source
best.ptSHA-256:4028152940f7c910f40192f46ede3b3f6c7129e5c76849c324d3564f8ac50198 - Ultralytics version recorded by the checkpoint:
8.3.227 - Architecture: YOLO11n instance segmentation, one
balloonclass, 1600-pixel inference size
The conversion copies all EMA model tensors without quantization or dtype changes.
PyTorch BatchNorm num_batches_tracked counters are omitted because they are not
model parameters and are not used during inference.
The original model and this conversion are distributed under Apache-2.0.