import numpy as np
import cv2

shape = (5000, 4000, 3)
random_numbers = np.random.uniform(150, 255, shape)

rows, cols, chans = random_numbers.shape

# Call into numpy using float32
random_numbers_32 = np.reshape(random_numbers, (rows * cols, chans)).astype(np.float32)
sum_float32 = random_numbers_32.sum(axis=0)
means_float32 = random_numbers_32.mean(axis=0)

print(f"Mean of random numbers (float32): {means_float32} | {sum_float32[0]:.6f} {sum_float32[1]:.6f} {sum_float32[2]:.6f}")

theo_32_limit = (2**32)/(shape[0]*shape[1])
print(f"Mean using theoretical 2^32 limit: {theo_32_limit}")

if(abs(theo_32_limit - means_float32[0]) < 0.0001):
    print(f"Match? YES")
else:
    print(f"Match? NO!")
    
# Call into numpy using float64
random_numbers_64 = np.reshape(random_numbers, (rows * cols, chans)).astype(np.float64)
sum_float64 = random_numbers_64.sum(axis=0)
means_float64 = random_numbers_64.mean(axis=0)

print(f"Mean of random numbers (float64): {means_float64} | {sum_float64[0]:.6f} {sum_float64[1]:.6f} {sum_float64[2]:.6f}")