A computer vision model could revolutionize trauma video review by automating the identification of key phases and procedures, improving trauma care quality.

• Achieved 98.3% frame-wise accuracy and high edit and F1 scores (up to 94.5%) for trauma resuscitation phases.
• Procedure detection accuracy surpassed 66% for x-rays and central line placements.

This technology can enhance surgical decision-making and promote wider adoption of trauma video review in practice.

• Interrater reliability among annotators was strong at 0.89.