[FEATURE_REQUEST] in evalution, add how many images are correctly identified

[jiata]

Description

Add evaluation how many images are correct:

[jiata]

Notes:

import numpy as np
e.eval_imgs["bbox"][3, 0, 0]

def eval_im():
    bbox = e.eval_imgs["bbox"]
    for i in range(len(bbox[0, 0, :])):
        im = []
        for j in range(len(bbox[:, 0, 0])):
            if bbox[j, 0, i]:
                dt_matches = bbox[j, 0, i]["dtMatches"]
                scores = bbox[j, 0, i]["dtScores"]
                for dt, score in zip(list(dt_matches[0]), scores):
                    im.append([dt, score])
        yield im
gen = eval_im()

next(gen)