RightWayUp: open 360° image rotation model in six ONNX / Core ML sizes

Ortus AI released RightWayUp, an Apache-2.0 licensed 360° image rotation model that predicts how far an image is rotated (0–359°) with a confidence score and abstains when it cannot determine which way is up. The model is published in six sizes, Pico to Max, as ONNX (FP32, FP16, INT8) and Core ML files on Hugging Face, with a tiers.json file supplying preprocessing and thresholds so the ONNX files can be called directly, plus ONNX Runtime Web builds for Pico. The training set includes 3,586 Blender renders released as a CC BY 4.0 dataset, and the developers say thermal and fisheye images are the cases they are least sure about.

Hi everyone. We needed to spot CCTV cameras that had been knocked or mounted crooked from a single frame, so we trained a model for it, and we’ve now opened it under Apache-2.0. RightWayUp predicts how far an image is rotated 0–359° with a confidence score, and abstains when it can’t tell which way is up. All six sizes, Pico to Max, are in one repo as ONNX FP32, FP16, INT8 and Core ML. You don’t need our package: tiers.json in the repo has the preprocessing and thresholds for every file, so you can call the ONNX files directly. Pico also has builds for ONNX Runtime Web. pip install rightwayup rightwayup fix photo.jpg The 3,586 Blender renders we trained on are up as a dataset too CC BY 4.0 , with the roll and camera details for every image. They’re a small part of the training data, added to make the model more robust on CCTV-style scenes. All the images used for training are in the manifests, with corresponding links, authors, and licences. Most useful to us: images where it’s confidently wrong. Thermal and fisheye are the ones we’re least sure about. Model: ortusai/rightwayup · Hugging Face https://huggingface.co/ortusai/rightwayup renders are in ortusai/rightwayup-renders next to it .