Returns hand gesture based on image with about 96% accuracy.
See https://www.kaggle.com/code/dima806/hand-gestures-image-detection-vit for more details.
Classification report:
precision recall f1-score support
call 0.9256 0.9752 0.9498 11825
dislike 0.9784 0.9862 0.9823 11826
fist 0.9833 0.9870 0.9851 11826
four 0.9140 0.9357 0.9247 11826
like 0.9761 0.9101 0.9420 11825
mute 0.9831 0.9964 0.9897 11826
ok 0.9586 0.9658 0.9622 11825
one 0.9708 0.9453 0.9579 11826
palm 0.9764 0.9637 0.9700 11826
peace 0.9187 0.9367 0.9276 11825
peace_inverted 0.9784 0.9748 0.9766 11826
rock 0.9439 0.9361 0.9400 11825
stop 0.9502 0.9723 0.9611 11825
stop_inverted 0.9828 0.9546 0.9685 11826
three 0.9135 0.9068 0.9101 11826
three2 0.9799 0.9670 0.9734 11826
two_up 0.9570 0.9766 0.9667 11826
two_up_inverted 0.9754 0.9703 0.9729 11825
accuracy 0.9589 212861
macro avg 0.9592 0.9589 0.9589 212861
weighted avg 0.9592 0.9589 0.9589 212861