raw/ - 120 RGB racetrack images.
labels/<sample_id>/target_track_mask.png - binary track-surface masks.
labels/<sample_id>/overlay.png - visual QA overlays.
labels/<sample_id>/label.json - full freeform labeler annotation payloads.
manifests/ - source generation manifest.
prompts/ - generation prompt bundle used for the synthetic set.
reports/ - dataset preview and generation report.
metadata.csv - ImageFolder-compatible table mapping each image to its mask, overlay and label JSON. The *_file_name columns let Hugging Face Dataset Viewer and datasets.load_dataset(...) decode image, mask and overlay as image columns.

Loading

from datasets import load_dataset

ds = load_dataset("RedRatInHat/racelinecalc-track-segmentation", split="train")
sample = ds[0]
image = sample["image"]
mask = sample["mask"]
overlay = sample["overlay"]

Dataset Contract

Each sample is a binary semantic segmentation pair:

input:  RGB top-down / satellite-like racetrack image
output: binary mask where positive pixels are drivable asphalt track surface

The dataset is intentionally narrow: it targets racetrack/asphalt surface extraction, not general road-scene segmentation.

Splits

The split used for the published baseline model is available in the model repository under training/split.json.

Related Model

The baseline mobile ONNX model trained on this dataset is published here:

RedRatInHat/racelinecalc-track-mask

Provenance

This dataset was generated and labeled during development of RaceLineCalc automatic track boundary detection. It is published as a full v1 research/portfolio dataset for reproducibility and external evaluation.