Speakers
Shicong Hu
(高能所)
Yong Huang
Description
In this report, we present our exploration of leveraging Vision Transformers to enhance the performance of WCDA. Using pre-trained ViT and Swin Transformer, we inject the hit's charge and time information into different color channels of the images and conduct a systematic study on the effects of patch number, image resolution, and other factors. Based on the above methodology, we achieve approximately a 1.6× sensitivity improvement in background rejection compared to traditional methods, along with certain gains in directional reconstruction.
Primary author
Shicong Hu
(高能所)
Co-author
Zhiguo Yao
(IHEP, Beijing, China)
