benchmark.py

import pandas as pd
import os
import cv2
import time

from recognizer import Recognizer

recog = Recognizer()

res = recog.det_model.val(data="C:\\Users\\USER\\Downloads\\Car license plate detection.v1i.yolov5pytorch\\data.yaml")
print(res.results_dict)

metrics = [0, 0]
path = "C:\\Users\\USER\\Downloads\\ocr-test"
started_at = time.time()
tests_count = 0
for k, name in enumerate(os.listdir(path)):
    for res in recog(cv2.imread(f'{path}\\{name}')):
##        print(res[0], name[:-4], res[0]==name[:-4], res[-1])
        if res[0]==name[:-4]:
            metrics[0] += 1
        metrics[1] += 1
##    if k%100==0:
##        print(metrics, f"round(metrics[0]*100/metrics[1], 2)%")
    tests_count += 1
print(f'Totally correct are {metrics[0]}/{metrics[1]}', f"{round(metrics[0]*100/metrics[1], 2)}%")
print(f'Finished recognizing {tests_count} pics in {round(time.time()-started_at, 2)} s')