Series: Oriented-Det

Technical notes on oriented-det: announcements, releases, training, evaluation, and parity work against research frameworks.

Start here

  1. Oriented-det is coming — motivation and design goals
  2. Oriented-Det v0.1.0 is out — install, docs, and getting started
  3. Oriented R-CNN detections for the 15 DOTA classes — qualitative gallery on the DOTA taxonomy
  4. Oriented object detection on macOS, in pure Python — hands-on inference with odet image-demo on Apple Silicon
  5. Zero-shot ships on Sentinel-2 — public checkpoint on a Copernicus tile
  6. Sliding-window inference on large aerial tiles — pad, tile, merge NMS with the Oriented R-CNN 1× Hub weight
  7. Rotated Faster R-CNN on DOTA without custom CUDA — ProbIoU, sampled rIoU, 74.42% official Task 1 vs MMRotate 73.40
  8. Oriented-Det v0.1.1 — ProbIoU packaged, MMRotate parity, harbor-scene demo
  9. Oriented-Det v0.2.0 — Rotated FCOS, decoded rIoU, four-family 1×/3× official Task 1 zoo
  10. Rotated FCOS vs Oriented R-CNN on macOS — Apple Silicon MPS latency, 1× L4 training wall, side-by-side demos
  11. A static demo of three oriented detectors — Oriented R-CNN 3×, Rotated Faster R-CNN 3×, and FCOS 3× on six optical scenes, in parity with MMRotate
  12. Apache 2.0 covers oriented-det. It does not cover DOTA or HRSC. — sovereignty of the stack versus research datasets; train on your own licensed imagery
  13. HRSC2016 recipes, trains, and results — native ship loader, three 3× Hub weights, held-out test 90.41% / 88.77% / 88.34%
  14. Lessons learned on DOTA — official Task 1, the leaky eval-val trap, and why the last mile was the box loss
  15. Oriented-Det v0.3.1 — four dataset loaders, HRSC Hub 3×, RetinaNet OBB, ONNX export
  16. FAIR1M fine-grained detection — 37 classes, convert/tile, Faster R-CNN 1× tiled-val 36.70%
  17. SSDD SAR ship finetune — optical DOTA → SAR, held-out test 90.34%
  18. HRSID SAR ship benchmark — larger SAR set, 78.55% rotated AP50 vs Wei HBB
  19. Deploy in Docker — Sanic Tile Geo Process, GeoJSON out, PyTorch in CUDA
  20. ONNX export without PyTorch — pre-NMS ONNX + ORT consumer stack
  21. Does MMRotate Faster R-CNN also mess up the bus lot? — official 1× Rotated Faster R-CNN is messy on the same DOTA bus tile as OrientedDet FRCNN (architecture, rare ~45° clutter); Oriented R-CNN is clean but heavy, which pushes toward FCOS
  22. Which oriented detector should you train? — Oriented R-CNN when the box must be tight (AP75); Faster R-CNN or FCOS when recall matters more, with FCOS for dense ~45° objects

Links