Retinal blood vessel segmentation in the eyeground

• The fundus retinal blood vessel segmentation application was developed for the Atlas 200 DK inference system, in partnership with the Nankai University, led by Professor Li Tao of the Intelligent Computing System Research Office.

  1. This project makes full use of the neural network computing power of the Atlas 200 DK system to segment the fundus vessels in real-time.

  2. The total inference time of 20 pictures is 761.8 milliseconds, and the average inference time of one image is 38 milliseconds.