Use Amazon Rekognition to detect world strength from captured images of Smash Bros SPECIAL. (part 2) DETAIL: https://michimani.net/post/aws-extract-value-of-smash-blos-sp-power-by-rekognition-2/

detect_smash_bros_sp_power_v2.py

import boto3
import cv2
import datetime
import hashlib
import json
import os
import re
import sys
import time

TMP_DIR = './tmp'
THUMBNAIL_DIR = './char_thumb'

reko = boto3.client('rekognition')

target_area = {
    'A': {'X': 0.57187, 'Y': 0.1875},
    'B': {'X': 0.9, 'Y': 0.1875},
    'C': {'X': 0.9, 'Y': 0.43125},
    'D': {'X': 0.57187, 'Y': 0.43125},
}

character_area = {
    'A': {'X': 911, 'Y': 474},
    'B': {'X': 959, 'Y': 474},
    'C': {'X': 959, 'Y': 522},
    'D': {'X': 911, 'Y': 522},
}


def get_datetime_from_image_path(image_path):
    datetime_part = re.sub(r'.*(\d{16}).*', r'\1', image_path)[:14]
    dt = datetime.datetime.strptime(datetime_part, '%Y%m%d%H%M%S')
    datetime_str = '{0:%Y-%m-%d %H:%M:%S}'.format(dt)
    return datetime_str


def get_character_hash(image_path):
    img = cv2.imread(image_path)

    trimed = img[
        character_area['A']['Y']: character_area['D']['Y'],
        character_area['A']['X']: character_area['C']['X']]

    trimed_bytes = trimed.tobytes()
    trimed_hash = hashlib.sha1(trimed_bytes).hexdigest()

    thumb_path = f'{THUMBNAIL_DIR}/{trimed_hash}.jpg'
    if os.path.exists(thumb_path) is False:
        cv2.imwrite(thumb_path, trimed)

    return trimed_hash


def get_bytes_of_image(image_path):
    img = cv2.imread(image_path)
    trimed = img[
        360: 720,
        640: 1280
    ]

    ts = str(time.time())
    trimed_tmp_path = f'{TMP_DIR}/{ts}.jpg'
    cv2.imwrite(trimed_tmp_path, trimed)

    with open(trimed_tmp_path, mode='rb') as img:
        image_bytes = img.read()

    os.remove(trimed_tmp_path)

    return image_bytes


def detect_text(image_path):
    image_bytes = get_bytes_of_image(image_path)
    return reko.detect_text(
        Image={
            'Bytes': image_bytes
        }
    )


def get_smash_power(image_path):
    smash_power = 0
    detect_res = detect_text(image_path)

    for detected in detect_res['TextDetections']:
        polygon = detected['Geometry']['Polygon']
        if polygon[0]['X'] > target_area['A']['X'] and polygon[0]['Y'] > target_area['A']['Y'] \
                and polygon[1]['X'] < target_area['B']['X'] and polygon[1]['Y'] > target_area['B']['Y'] \
                and polygon[2]['X'] < target_area['C']['X'] and polygon[2]['Y'] < target_area['C']['Y'] \
                and polygon[3]['X'] > target_area['D']['X'] and polygon[3]['Y'] < target_area['D']['Y']:
            smash_power = int(detected['DetectedText'].replace(',', ''))

    return smash_power


def generate_smash_power_item(image_path):
    # Datetime of data
    datetime_str = get_datetime_from_image_path(image_path)

    # Get Smash Power via Amazon Rekognition
    smash_power = get_smash_power(image_path)

    # Get character hash (and Generate character thumbnail)
    character_hash = get_character_hash(image_path)

    return {
        'character_hash': character_hash,
        'capture_datetime': datetime_str,
        'power': smash_power
    }


if __name__ == '__main__':
    image_path = sys.argv[1]
    smash_power_item = generate_smash_power_item(image_path)
    print(json.dumps(smash_power_item, indent=2, ensure_ascii=False))