: Oriented-Det
Technical notes on oriented-det: announcements, releases, training, evaluation, and parity work against research frameworks.
Start here
- Oriented-det is coming — motivation and design goals
- Oriented-Det v0.1.0 is out — install, docs, and getting started
- Oriented object detection on macOS, in pure Python — hands-on inference with
odet image-demoon Apple Silicon - Oriented R-CNN detections for the 15 DOTA classes — qualitative gallery on the DOTA taxonomy
- Zero-shot ships on Sentinel-2 — public checkpoint on a Copernicus tile
- Announcing the Oriented R-CNN 3× pretrained model — first DOTA le90 zoo chapter
- Rotated Faster R-CNN on DOTA without custom CUDA — ProbIoU, sampled rIoU, 83.42% eval-val
- Oriented-Det v0.1.1 — ProbIoU packaged, MMRotate parity, harbor-scene demo
- Oriented-Det v0.2.0 — Rotated FCOS, decoded rIoU, four-family zoo
- Rotated FCOS vs Oriented R-CNN on macOS — Apple Silicon MPS latency, L4 training wall (~5.7×), side-by-side demos
- A static demo of three oriented detectors — Rotated Faster R-CNN, FCOS, and Oriented R-CNN on seven optical scenes, in parity with MMRotate
Links
- GitHub · PyPI · Documentation
-
A static demo of three oriented detectors on optical satellite imagery
A browser demo of Oriented-Det’s three strongest DOTA 3× checkpoints — Rotated Faster R-CNN, Rotated FCOS, and Oriented R-CNN — on seven optical satellite scenes, with a side by side comparison that lands in parity with state of the art frameworks.
-
Rotated FCOS vs Oriented R-CNN on macOS
Hands-on Apple Silicon comparison of Hub 3× DOTA checkpoints — Rotated FCOS (82.32% eval-val) vs Oriented R-CNN (79.40%) — MPS latency, training wall on NVIDIA L4 (~5.7× faster for FCOS), score thresholds, and side-by-side detections.
-
Oriented-Det v0.2.0 — Rotated FCOS, decoded rIoU, and a four-family zoo
Oriented-det v0.2.0 is on PyPI — Rotated FCOS joins the zoo as the balanced one-stage detector, with a decoded rIoU 3× Hub checkpoint at 82.32% eval-val mAP50, and the same Apache 2.0 stack.
-
Oriented-Det v0.1.1 — ProbIoU, MMRotate parity, and the updated zoo
Oriented-det v0.1.1 is on PyPI — ProbIoU ROI regression, MMRotate-aligned training fixes, published eval reports, a refreshed DOTA le90 zoo led by Rotated Faster R-CNN 3× at 83.42% eval-val mAP50 (~7× faster inference than Oriented R-CNN 3×), and a hands-on harbor-scene demo.
-
Rotated Faster R-CNN on DOTA without custom CUDA: sampled rIoU, ProbIoU, and an 83.4% checkpoint
Why OrientedDet avoids MMRotate's exact CUDA IoU kernels, how ProbIoU trains oriented boxes in pure PyTorch, and results from our published 3× DOTA le90 model.
-
Announcing the Oriented R-CNN 3× pretrained model
The OrientedDet DOTA le90 pretrained zoo is complete. Oriented R-CNN 3× reaches 79.4% eval-val mAP50; Rotated Faster R-CNN 3× with ProbIoU now leads the zoo at 83.4%.
-
Zero-shot Ship detection on a Copernicus Sentinel-2 tile with Oriented R-CNN
A practical zero-shot experiment with a DOTA-pretrained Oriented R-CNN model on a 10 m Copernicus Sentinel-2 tile, including zoomed sliding-window inference and ship-only filtering.
-
Oriented object detection on macOS, in pure Python
Run Oriented R-CNN on Apple Silicon with pip, uv, and the odet CLI — no CUDA toolchain, from demo.jpg to result.jpg in minutes.
-
Oriented R-CNN detections for the 15 DOTA classes
A visual tour of Oriented R-CNN detections across the 15 original DOTA v1.0 aerial object classes.
-
Oriented-Det v0.1.0 is out — sovereign oriented object detection for EO
Oriented-det v0.1.0 is on PyPI under Apache 2.0 — a lightweight PyTorch library for rotated object detection in aerial and satellite imagery, with DOTA baselines and pretrained weights.
-
Oriented-det is coming: sovereign oriented detection for EO
A sovereign, Apache-licensed oriented detection stack for EO, targeting an official release in June 2026.