Convolutional nets or Vision Transformers?

my team is tasked with creating a computer vision feature which will need to be able to do the following things in production:

1. Object Detection
2. Face Recognition
3. Gaze tracking

it'd be hugely beneficial if any of you guys could advice on which model would be the most optimal to get started with?

advice computer computer vision computervision convolutional detection face face recognition feature object production recognition team things tracking transformers vision vision transformers will you