CTU Robotics Vision Toolkit

Computer-vision toolkit for robotics workflows with ArUco localization, panel processing, and MobileNetV3 training/inference.

This toolkit is based on Improutils research-lab work, merged into a unified core utility API.

Installation

Quick install:

pip install .

Optional extras:

pip install .[ml]
pip install .[panel]
pip install .[dev]

Full setup (all extras):

pip install .[all]

You can also use requirements-all.txt for a full pinned-style environment.

Public API

import ctu_robotics_vision_toolkit as crvt

from ctu_robotics_vision_toolkit.core import ensure_uint8_image, crop_by_box
from ctu_robotics_vision_toolkit.aruco import ArucoDetector
from ctu_robotics_vision_toolkit.models import ONNXInference
from ctu_robotics_vision_toolkit.pipelines import (
    MobileNetV3TrainingPipeline,
    PanelCropExtractionPipeline,
    PanelEndToEndPipeline,
    PanelStateInferencePipeline,
)

CLI

crvt -h
crvt dataset-extract -h

Build Wheel

python -m build

The wheel and source distribution are created in dist/.

Docs

Markdown documentation lives in docs/.