Chronic rhinosinusitis (CRS) is a persistent inflammatory disease affecting 1 in 8 adults in the US. CRS profoundly affects health-related quality-of-life and is commonly treated with endoscopic sinus surgery (ESS). ESS fails to induce durable symptom improvements in 25% of CRS patients and subsequent revision ESS is needed in 15-46% of cases due to persistent or recurrent symptoms after incomplete surgical dissection. Revision ESS has significantly lower success rates than primary ESS and independently predicts poorer clinical outcomes. Complete surgical dissection, ideally during the first ESS, is less costly and is most likely to improve symptoms. Therefore, tools that enable complete ESS are critical. Image-guided surgery (IGS) can facilitate surgical dissection in ESS by mapping the location of surgical instruments to preoperative computed tomography (CT) images. However, IGS systems have significant limitations: 1) high costs, driving lower IGS usage, especially in underserved groups; 2) loss of tracking accuracy during ESS; and 3) disparity between the static CT images and reality as ESS progresses. We aim to establish a computer vision-based navigation system that will: 1) greatly reduce the cost of surgical navigation by eliminating expensive IGS hardware, thereby democratizing the use of navigation; 2) achieve consistent submillimeter accuracy during ESS; and 3) enhance surgical completeness to drive better clinical outcomes in CRS patients. The objective of this proposal is to develop a low-cost vision-based navigation system that reflects dynamic changes in surgical anatomy on the CT, maps critical anatomy from the CT onto a 3D reconstruction of the surgical field, and continuously tracks surgical instruments that are in view, aggregating these data in real-time for visualization. The central hypothesis is that vision-based navigation will maintain accuracy during ESS and will display critical anatomic structures and up- to-date anatomic changes to the s