thirdeyelabs 0.1.0__tar.gz

thirdeyelabs-0.1.0/PKG-INFO

@@ -0,0 +1,69 @@
+Metadata-Version: 2.5
+Name: thirdeyelabs
+Version: 0.1.0
+Summary: Open road data for autonomous driving on unstructured Indian roads.
+Project-URL: Homepage, https://thirdeyelabs.ai
+Project-URL: Dataset, https://huggingface.co/datasets/thirdeyelabs/indian-road-dataset
+Author: Third Eye Labs
+License: Apache-2.0
+Keywords: adas,autonomous-driving,computer-vision,dataset,india,object-detection
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Scientific/Engineering :: Image Recognition
+Requires-Python: >=3.9
+Provides-Extra: hf
+Requires-Dist: datasets>=2.14.0; extra == 'hf'
+Description-Content-Type: text/markdown
+
+# thirdeyelabs
+
+Open road data for autonomous driving on unstructured Indian roads.
+
+Every major AV dataset (Waymo Open, nuScenes, Argoverse, PandaSet) was captured
+on structured Western roads. This one was not.
+
+```bash
+pip install thirdeyelabs[hf]
+```
+
+```python
+import thirdeyelabs as tel
+
+tel.info()
+# {'frames': 646014, 'clips': 8441, 'detections': 6896202, ...}
+
+tel.INDIA_SPECIFIC
+# ['autorickshaw', 'animal', 'vehicle_fallback']
+
+ds = tel.load_dataset()
+```
+
+## The dataset
+
+| | |
+|---|---|
+| Frames | 646,014 across 8,441 clips |
+| Detections | 6,896,202 2D boxes |
+| Segmentation | 1,290,463 masks |
+| Classes | 12, BDD100K schema plus India-specific categories |
+| Region | Delhi NCR, India |
+| Sensor | Monocular dashcam, GPS per clip |
+| License | CC BY 4.0 |
+
+Three of the twelve categories (`autorickshaw`, `animal`, `vehicle_fallback`)
+have no equivalent in COCO or BDD100K, so a detector trained on those
+benchmarks cannot emit them at all.
+
+Dataset: https://huggingface.co/datasets/thirdeyelabs/indian-road-dataset
+
+## Status
+
+Early. This release provides dataset access helpers and the taxonomy.
+Simulation environments and benchmark suites are in development.
+
+## License
+
+Package: Apache-2.0. Dataset: CC BY 4.0.

thirdeyelabs-0.1.0/README.md

@@ -0,0 +1,49 @@
+# thirdeyelabs
+
+Open road data for autonomous driving on unstructured Indian roads.
+
+Every major AV dataset (Waymo Open, nuScenes, Argoverse, PandaSet) was captured
+on structured Western roads. This one was not.
+
+```bash
+pip install thirdeyelabs[hf]
+```
+
+```python
+import thirdeyelabs as tel
+
+tel.info()
+# {'frames': 646014, 'clips': 8441, 'detections': 6896202, ...}
+
+tel.INDIA_SPECIFIC
+# ['autorickshaw', 'animal', 'vehicle_fallback']
+
+ds = tel.load_dataset()
+```
+
+## The dataset
+
+| | |
+|---|---|
+| Frames | 646,014 across 8,441 clips |
+| Detections | 6,896,202 2D boxes |
+| Segmentation | 1,290,463 masks |
+| Classes | 12, BDD100K schema plus India-specific categories |
+| Region | Delhi NCR, India |
+| Sensor | Monocular dashcam, GPS per clip |
+| License | CC BY 4.0 |
+
+Three of the twelve categories (`autorickshaw`, `animal`, `vehicle_fallback`)
+have no equivalent in COCO or BDD100K, so a detector trained on those
+benchmarks cannot emit them at all.
+
+Dataset: https://huggingface.co/datasets/thirdeyelabs/indian-road-dataset
+
+## Status
+
+Early. This release provides dataset access helpers and the taxonomy.
+Simulation environments and benchmark suites are in development.
+
+## License
+
+Package: Apache-2.0. Dataset: CC BY 4.0.

thirdeyelabs-0.1.0/pyproject.toml

@@ -0,0 +1,35 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "thirdeyelabs"
+version = "0.1.0"
+description = "Open road data for autonomous driving on unstructured Indian roads."
+readme = "README.md"
+requires-python = ">=3.9"
+license = { text = "Apache-2.0" }
+authors = [{ name = "Third Eye Labs" }]
+keywords = [
+  "autonomous-driving", "computer-vision", "dataset",
+  "object-detection", "india", "adas",
+]
+classifiers = [
+  "Development Status :: 3 - Alpha",
+  "Intended Audience :: Science/Research",
+  "License :: OSI Approved :: Apache Software License",
+  "Programming Language :: Python :: 3",
+  "Topic :: Scientific/Engineering :: Artificial Intelligence",
+  "Topic :: Scientific/Engineering :: Image Recognition",
+]
+dependencies = []
+
+[project.optional-dependencies]
+hf = ["datasets>=2.14.0"]
+
+[project.urls]
+Homepage = "https://thirdeyelabs.ai"
+Dataset = "https://huggingface.co/datasets/thirdeyelabs/indian-road-dataset"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/thirdeyelabs"]

thirdeyelabs-0.1.0/src/thirdeyelabs/__init__.py

@@ -0,0 +1,33 @@
+"""Third Eye Labs: open road data for autonomous driving on Indian roads.
+
+The dataset is published under CC BY 4.0 at
+https://huggingface.co/datasets/thirdeyelabs/indian-road-dataset
+
+    >>> import thirdeyelabs as tel
+    >>> tel.info()
+    >>> tel.CLASSES
+    >>> ds = tel.load_dataset()        # requires: pip install thirdeyelabs[hf]
+"""
+
+from .dataset import (
+    CLASSES,
+    DATASET_ID,
+    DATASET_URL,
+    INDIA_SPECIFIC,
+    STATS,
+    info,
+    load_dataset,
+)
+
+__version__ = "0.1.0"
+
+__all__ = [
+    "CLASSES",
+    "DATASET_ID",
+    "DATASET_URL",
+    "INDIA_SPECIFIC",
+    "STATS",
+    "info",
+    "load_dataset",
+    "__version__",
+]

thirdeyelabs-0.1.0/src/thirdeyelabs/dataset.py

@@ -0,0 +1,68 @@
+"""Access helpers for the Third Eye Labs Indian road dataset."""
+
+from __future__ import annotations
+
+from typing import Any, Dict, List
+
+DATASET_ID = "thirdeyelabs/indian-road-dataset"
+DATASET_URL = f"https://huggingface.co/datasets/{DATASET_ID}"
+
+#: Detection categories. The taxonomy follows the BDD100K schema, extended with
+#: the classes that matter on Indian roads and have no Western equivalent.
+CLASSES: List[str] = [
+    "person",
+    "rider",
+    "car",
+    "truck",
+    "bus",
+    "motorcycle",
+    "bicycle",
+    "autorickshaw",
+    "animal",
+    "vehicle_fallback",
+    "traffic_light",
+    "traffic_sign",
+]
+
+#: Categories absent from COCO and BDD100K, so no detector trained on those
+#: benchmarks can emit them at all.
+INDIA_SPECIFIC: List[str] = ["autorickshaw", "animal", "vehicle_fallback"]
+
+STATS: Dict[str, Any] = {
+    "frames": 646_014,
+    "clips": 8_441,
+    "detections": 6_896_202,
+    "segmentation_masks": 1_290_463,
+    "classes": len(CLASSES),
+    "region": "Delhi NCR, India",
+    "sensor": "monocular dashcam",
+    "annotation_schema": "BDD100K",
+    "license": "CC BY 4.0",
+}
+
+
+def info() -> Dict[str, Any]:
+    """Return dataset statistics and where to get it."""
+    return {**STATS, "dataset_id": DATASET_ID, "url": DATASET_URL}
+
+
+def load_dataset(split: str = "train", **kwargs: Any):
+    """Load the dataset from the Hugging Face Hub.
+
+    Requires the optional dependency::
+
+        pip install thirdeyelabs[hf]
+
+    Args:
+        split: Dataset split to load.
+        **kwargs: Forwarded to ``datasets.load_dataset``.
+    """
+    try:
+        from datasets import load_dataset as _hf_load_dataset
+    except ImportError as exc:  # pragma: no cover
+        raise ImportError(
+            "The 'datasets' package is required. "
+            "Install it with: pip install thirdeyelabs[hf]"
+        ) from exc
+
+    return _hf_load_dataset(DATASET_ID, split=split, **kwargs)