Report submission is open¶
All paper submissions should be sent directly to the email addresses of the organizers (jiang841@purdue.edu or zhuoy@purdue.edu) or the submission platform: https://easychair.org/my/conference?conf=t3challenge. If the files are too large and cannot be attached to the email, please send the cloud storage link. The files will be downloaded after the deadline.
Paper template: https://www.springer.com/gp/computer-science/lncs/conference-proceedings-guidelines
The primary goal of The Trauma THOMPSON Challenge is to find the best algorithms for automatic action recognition and prediction using computer vision from first-person view in the medical domain (refer to egocentric datasets of medical procedures). We offer the first egocentric view dataset of life-saving intervention (LSI) procedures with detailed annotations by medical professionals. We have collected over 200 procedure videos with environment, simulator, and type variability. Based on this dataset, the challenge we propose involves multiple tasks to encourage participants across the globe to design impactful algorithms with applications to medicine. The envisioned algorithms include action recognition, action anticipation, procedure recognition, and visual question answering (VQA).
Five Procedures: ¶
- Tube thoracostomy
- Needle thoracostomy
- Interosseous insertion
- Tracheostomy
- Tourniquet application
Four Tasks: ¶
● Track 1
o Task 1: Action Recognition
o Task 2: Action Anticipation
o Task 3: Procedure Recognition
● Track 2
o Task 1: Visual Question Answering