Returns military aircraft type given cropped image with about 76% accuracy.
See https://www.kaggle.com/code/dima806/military-aircraft-detection-vit for more details.
Classification report:
precision recall f1-score support
A10 0.8156 0.8889 0.8507 612
A400M 0.8641 0.7794 0.8196 612
AG600 0.9452 0.9592 0.9521 612
AV8B 0.4991 0.9003 0.6422 612
B1 0.8358 0.8154 0.8255 612
B2 0.8924 0.9624 0.9261 612
B52 0.9354 0.7337 0.8223 612
Be200 0.8491 0.8922 0.8701 612
C130 0.9104 0.4984 0.6441 612
C17 0.8045 0.5310 0.6398 612
C2 0.7765 0.8971 0.8324 612
C5 0.6826 0.7239 0.7026 612
E2 0.8866 0.9706 0.9267 612
E7 0.8045 0.9951 0.8897 612
EF2000 0.7348 0.2173 0.3354 612
F117 0.8298 0.9722 0.8954 612
F14 0.6531 0.7075 0.6792 612
F15 0.6058 0.2059 0.3073 612
F16 0.5390 0.2598 0.3506 612
F18 0.5905 0.5866 0.5885 612
F22 0.6273 0.7369 0.6777 612
F35 0.5764 0.6536 0.6126 612
F4 0.6749 0.3562 0.4663 612
H6 0.9245 0.9199 0.9222 612
J10 0.5846 0.6209 0.6022 612
J20 0.8477 0.5458 0.6640 612
JAS39 0.5615 0.4771 0.5159 612
JF17 0.4866 0.9477 0.6430 612
KC135 0.7706 0.9167 0.8373 612
MQ9 0.8618 0.9167 0.8884 612
Mig31 0.7900 0.6699 0.7250 612
Mirage2000 0.8333 0.3758 0.5180 612
P3 0.7997 0.9265 0.8584 612
RQ4 0.8433 0.9412 0.8896 612
Rafale 0.4622 0.5801 0.5145 612
SR71 0.9030 0.9281 0.9154 612
Su24 0.7059 0.6275 0.6644 612
Su25 0.8569 0.8023 0.8287 612
Su34 0.8042 0.8856 0.8429 612
Su57 0.7061 0.8873 0.7864 612
Tornado 0.6004 0.5523 0.5753 612
Tu160 0.8468 0.9575 0.8988 612
Tu22M 0.7758 0.9444 0.8519 612
Tu95 0.9271 0.8938 0.9101 612
U2 0.9094 0.7712 0.8347 612
US2 0.7837 0.9118 0.8429 612
V22 0.8788 0.9363 0.9066 612
Vulcan 0.8971 0.8693 0.8830 612
XB70 0.8857 1.0000 0.9394 612
YF23 0.8673 0.9935 0.9261 612
accuracy 0.7608 30600
macro avg 0.7690 0.7608 0.7488 30600
weighted avg 0.7690 0.7608 0.7488 30600