Omatrack telemetry reader 1.0.0

Upstream component: PyTorch Image Models (timm), MobileNetV3-small
Copyright 2019 Ross Wightman
Licensed under the Apache License, Version 2.0.
Full upstream license: LICENSE-APACHE-2.0-UPSTREAM.txt
Source: https://github.com/huggingface/pytorch-image-models
Pretrained model: https://huggingface.co/timm/mobilenetv3_small_100.lamb_in1k

Modification notice: this is not the original ImageNet classifier. The model
retains the pretrained stem and early backbone blocks, adds a crop encoder,
digit/CTC, visible-fill and digit-count heads, fine-tunes for gauge reading,
and exports the resulting task model to ONNX. No new training was performed
for this publication. Private training data and original checkpoints are not
included.

Omatrack-derived example source code
Copyright (c) 2026 Omatrack contributors
MIT License: LICENSE-MIT-OMATRACK-CODE.txt
The read_frame.py and test_read_frame.py examples adapt the public Omatrack
reader conventions and contain no private image or recording fixtures.
Source: https://github.com/tobi/omatrack

The license for the new task-specific model weights has not yet been assigned.
The upstream Apache-2.0 and example-code MIT notices are scoped to their
respective components; neither is a blanket license grant for the new model.
No rights to private footage, screenshots, labels or datasets are granted.
Third-party names identify compatibility/provenance, not endorsement.