Thanks for the further performance pointers! Is (1) primarily avoided by trimming off as much of the skins as possible (without damaging the follicles) and (2) by inserting them so that the hairs grow out closer to perpendicular to the scalp (rather than tangential). I speculate there is a lot of room for automation during the “filtering” the grafts based on follicle counts. I’ve read that in some areas of visual-diagnostic medicine, computer models are able to classify images more accurately than medical professionals (on a nonmedical image classification task, a convolutional neural net I built was more accurate than me in a small labeled test sample). But perhaps even more important is that it can reduce the time needed for the procedure, and therefore, also reduce the costs (which allows more people to get a transplant).
WE have monthly open hose events where our former patients come in to talk about their experience. Most people like you who come are skeptical thinking that can tell who had a transplant and they are always pleasantly surprised that they can’t tell. Good transplants can’t be detected. Look at Elan Musk before and after his transplant, a good example of a good job with actor Joe Penny (https://baldingblog.com/actor-joe-penny-shows-off-his-repaired-hairline-with-photos/).