Surgical video understanding baselines
Collection
LoRA adapters, supervised CNN baselines, and LemonFM linear probes trained for the SDSC x Chicago Booth surgical video understanding leaderboard. • 16 items • Updated
How to use skblv/yolov12m-surgvu-instruments with ultralytics:
from ultralytics import YOLOvv12
model = YOLOvv12.from_pretrained("skblv/yolov12m-surgvu-instruments")
source = 'http://images.cocodataset.org/val2017/000000039769.jpg'
model.predict(source=source, save=True)Supervised YOLOv12-m detector trained for the 17 SurgVU instruments, then scored as multi-label presence.
This is the “YOLOv12-m” row on the Instruments tab of the SDSC × Chicago Booth surgical video understanding leaderboard.
Full SurgVU validation split (95% bootstrap CI):
| Metric | Value |
|---|---|
| Exact match | 51.75% (50.97–52.51) |
| Micro-averaged F1 | 75.34% (74.87–75.82) |
Derived from Ultralytics YOLO weights; distributed under AGPL-3.0.
Research baseline only. Not a medical device.