RGB search with numba.cuda - 10 x faster than numpy pip install locate-pixelcolor-numbacuda from locate_pixelcolor_numbacuda import search_colors from a_cv_imwrite_imread_plus import open_image_in_cv import numpy as np colors=[(66, 71, 69),(62, 67, 65),(144, 155, 153),(52, 57, 55),(127, 138, 136),(53, 58, 56),(51, 56, 54),(32, 27, 18),(24, 17, 8),] https://www.pexels.com/pt-br/foto/foto-da-raposa-sentada-no-chao-2295744/ picnp = open_image_in_cv('pexels-alex-andrews-2295744.jpg',channels_in_output=3) coords=search_colors(pic=picnp,colors=colors,threadsperblock=(18, 18,3),dtypetouse = np.int32) print(coords) %timeit search_colors(pic=picnp,colors=colors,threadsperblock=(18, 18,3),dtypetouse = np.int32) # [[[ 19 14] # [ 19 14] # [ 11 17] # ... # [6613 4524] # [6614 4524] # [6615 4524]]] # 135 ms ± 3.5 ms per loop (mean ± std. dev. of 7 runs, 10 loops each) # More benchmarks: https://github.com/hansalemaos/locate_pixelcolor_cpp