aiqbench analysis
This article introduces woma, a self-supervised foundation model designed for real-time gastrointestinal endoscopy that achieves high-precision detection of polyps and lesions without requiring labeled training data. The research demonstrates a robust production pipeline using custom Vulkan-based kernels to deliver inference speeds exceeding 100 fps on standard workstation GPUs, outperforming major frameworks like PyTorch and TensorRT in specific regimes.
woma: a real-time foundation model and its fine-tuned models for endoscopy
arXiv:2609.15130v1 Announce Type: cross Abstract: woma is a real-time foundation model for gastrointestinal endoscopy: a network trained without labels on about a million endoscopy frames, from which task models are fine-tuned. We contribute a systematic design for production. Requirements and pass marks were fixed before any run, eight candidates screened under pre-registered rules, self-supervised training taken to a stopping rule, then fine-tuning and deployment optimisation, all on one self-contained library, numbat. We also contribute woma itself with two fine-tuned models, every outcome reported met or missed. Our colonoscopy model finds and outlines polyps, names which colon segment is in view, suggests polyp type and grades bowel preparation. Our gastroscopy model names a station out of 22 protocol sites, flags and outlines lesions, and names one of seven findings. Every number was read on data never seen in training, and shipped weights were chosen on that record. In colonoscopy, 96% of polyps in a six-hospital PolypGen set are found at precision >=0.85, and 19 of 19 polyps across fifteen full REAL-Colon videos at 1.6 false alarms per procedure. In gastroscopy, landmark region is named correctly on 92% of frames from unseen patients, and 37 of 39 held-out neoplasia frames are flagged at specificity 0.91. On one workstation GPU every task runs over 1080p video at about 100 frames per second, faster than PyTorch, ONNX Runtime and TensorRT in all four precision regimes tested. TensorRT comes closest: one pass of our foundation model takes it 3 to 27% longer than ours, and we deliver 6 to 31% more frames per second from frame to results. A second build links no vendor library at all -- our own kernels over Vulkan -- so a site deploys two files and needs no toolkit, no cuDNN and no framework; in f32 it beats the CUDA build on the same card.
Abstract:woma is a real-time foundation model for gastrointestinal endoscopy: a network trained without labels on about a million endoscopy frames, from which task models are fine-tuned. We contribute a systematic design for production. Requirements and pass marks were fixed before any run, eight candidates screened under pre-registered rules, self-supervised training taken to a stopping rule, then fine-tuning and deployment optimisation, all on one self-contained library, numbat. We also contribute woma itself with two fine-tuned models, every outcome reported met or missed. Our colonoscopy model finds and outlines polyps, names which colon segment is in view, suggests polyp type and grades bowel preparation. Our gastroscopy model names a station out of 22 protocol sites, flags and outlines lesions, and names one of seven findings. Every number was read on data never seen in training, and shipped weights were chosen on that record. In colonoscopy, 96% of polyps in a six-hospital PolypGen set are found at precision >=0.85, and 19 of 19 polyps across fifteen full REAL-Colon videos at 1.6 false alarms per procedure. In gastroscopy, landmark region is named correctly on 92% of frames from unseen patients, and 37 of 39 held-out neoplasia frames are flagged at specificity 0.91. On one workstation GPU every task runs over 1080p video at about 100 frames per second, faster than PyTorch, ONNX Runtime and TensorRT in all four precision regimes tested. TensorRT comes closest: one pass of our foundation model takes it 3 to 27% longer than ours, and we deliver 6 to 31% more frames per second from frame to results. A second build links no vendor library at all -- our own kernels over Vulkan -- so a site deploys two files and needs no toolkit, no cuDNN and no framework; in f32 it beats the CUDA build on the same card.
28 pages, 10 figures (8 in the main text, 2 supplementary), 13 tables (8 in the main text, 5 supplementary). Preprint. Models and run records are available from the corresponding author
Software Engineering (cs.SE); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
D.2.1; D.2.11; I.2.6; I.4.6; J.3
arXiv:2609.15130 [cs.SE]
(or arXiv:2609.15130v1 [cs.SE] for this version)
https://doi.org/10.48550/arXiv.2609.15130 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 363 words
Related coverage
NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI
NVIDIA · May 21, 2026
Building a Foundation Stack for General-Purpose Robots
IEEE Spectrum · Jul 13, 2026
Presentation: Engineering AI for Creativity and Curiosity on Mobile
InfoQ · Jul 21, 2026
The generative AI customization spectrum: From prompt engineering to custom models on AWS
AWS · Sep 14, 2026