Rotoclear AI Chatbot ↗
A retrieval-augmented assistant connecting product knowledge with natural-language questions.
Organisation repository / AI & vision
Hailo Training Pipeline
A model preparation and optimisation pipeline for Hailo AI accelerators.
Discuss the work ↗Hailo Training PipelineA visual introduction
HUMAN + MACHINE
Look closer.
Synthetic inspection scene
01 / THE CONTEXTA closer look at
A closer look at
the work.
A model preparation and optimisation pipeline for Hailo AI accelerators.
TOOLS & TECHNOLOGIES
PythonHailoONNXMLflow
This repository belongs to an organisation I work with. Inclusion describes the project landscape and is not a claim of individual authorship.
SYSTEM MAPSelect a component to explore ↙
From dataset to accelerator
COCO
Prepare labelled training data.
Simplified architecture · illustrative, not a product screenshot
02 / INSIDE THE PROJECT
Behind the diagram.
Technical notes, when you want to go deeper.
01Implementation
The pipeline connects COCO dataset preparation, PyTorch training and model optimisation with Hailo compilation. MLflow records experiments while ONNX, HAR and HEF represent different stages of the model lifecycle.
02Key decision
Training artefacts and device-ready binaries are kept conceptually distinct. A model that trains successfully still needs conversion and accelerator-specific compilation before it can run on the target.
03Current scope
The repository describes the preparation and compilation workflow, with a GPU recommended for training. No accuracy, throughput or hardware acceptance figures are asserted here.
REPOSITORY NOTES
A closer look.
Source code is private. The portfolio contains a high-level description, without internal files or customer data.
Discuss the work ↗- Repository
- hailo-train-pipeline
- Editorial basis
- Repository documentation
- Default branch
- main