Topic Coverage¶

DL-Hub now keeps the broad community topic pool in a machine-checkable coverage registry instead of relying on README claims alone.

The registry lives in dlhub/topic_coverage.py and maps every requested topic to one or more concrete code artifacts. It covers three shapes of topic:

• model or task directions, such as detection, segmentation, 3D point cloud, video understanding, deraining, OCR, and geo-localization
• research workflow streams, such as paper digests, resource sharing, open-source projects, surveys, datasets, and tutorials
• cross-cutting method/framework topics, such as NAS, AutoML, pruning, distillation, SLAM, metaverse scene assets, PyTorch, TensorFlow, MXNet, TensorRT, OpenCV, NumPy, and Python

The runtime helpers are:

from dlhub.topic_coverage import coverage_report, describe_topic, validate_topic_coverage

report = coverage_report()
validate_topic_coverage()
print(describe_topic("目标检测").primary_artifact.module)

Additional lightweight surfaces:

• dlhub/research_streams.py for paper/resource/survey/dataset/tutorial topics
• dlhub/framework_adapters.py for optional framework probes without importing heavyweight packages
• dlhub/method_kits.py for runnable NAS/AutoML, pruning, distillation, SLAM, capsule-routing, and scene-asset utilities

Regression coverage is in tests/test_topic_coverage.py. The test checks that every topic in the requested pool has an existing path and an importable module, then exercises lookup, stream, framework, and method-kit behavior.

python -m pytest tests/test_topic_coverage.py

闭环层 / Coverage layers¶