Modelpublic

skblv/yolo11m-cls-sarrarp50-gesture

sourceHugging Faceagpl-3.0updated 26d agoView on Hugging Face

0likes25downloads

Model Card

YOLO11m-cls — SAR-RARP50 suturing gesture recognition

Supervised YOLO11m-cls baseline recognizing the current suturing action on robot-assisted radical prostatectomy frames from SAR-RARP50.

Trained as a baseline for the SDSC × Chicago Booth surgical video understanding leaderboard (Skill assessment tab, used as a skill proxy).

Prompt example

This closed-set example mirrors the leaderboard format, not a text-input API for this checkpoint.

text

[surgical frame]

What suturing action is being performed in this frame?
Choose one label.

- Other
- Picking Up The Needle
- Positioning The Needle Tip
- Pushing The Needle Through The Tissue
- Pulling The Needle Out Of The Tissue
- Tying A Knot
- Cutting The Suture
- Returning Or Dropping The Needle

Model

—Ultralytics yolo11m-cls.pt fine-tuned for 8-way single-label gesture classification
—224×224 inputs, batch 32, up to 100 epochs with patience 15, seed 42, standard color/geometric augmentation
—Full training code in s73_sarrarp50_supervised.py; training curves in loss_curve.csv

Evaluation

Full 636-frame validation split (1 Hz frames from held-out operations; 95% bootstrap CI):

Metric	Value
Exact-match accuracy	53.1% (49.4–57.2)

Best result on the Skill assessment leaderboard as of Aug 2026; see the leaderboard.

Usage

python

from ultralytics import YOLO

model = YOLO("best.pt")
result = model("frame.jpg")[0]
print(result.names[result.probs.top1])

License note

Derived from Ultralytics YOLO11 weights; this checkpoint is therefore distributed under AGPL-3.0.

References

—Skobelev, K., Fithian, E., Baranovski, Y., et al. A Comparative Study in Surgical AI: Potential and Limitations of Data, Compute, and Scaling. arXiv:2603.27341, 2026.
—Dataset: Psychogyios, D., Colleoni, E., Van Amsterdam, B., et al. SAR-RARP50: Segmentation of surgical instrumentation and Action Recognition on Robot-Assisted Radical Prostatectomy Challenge. arXiv:2401.00496, 2024.

Limitations

Research baseline only. Not a medical device. Single-frame gesture recognition misses temporal cues; this is a proxy task, not an OSATS/GRS skill score.