仮保存

import boto3
import cv2
import numpy as np
import matplotlib.pyplot as plt
from io import BytesIO
from botocore.exceptions import NoCredentialsError
import tempfile
import os

# AWSの認証情報を設定
aws_access_key_id = 'YOUR_AWS_ACCESS_KEY_ID'
aws_secret_access_key = 'YOUR_AWS_SECRET_ACCESS_KEY'
region_name = 'YOUR_AWS_REGION'

# S3クライアントを作成
s3 = boto3.client(
   's3',
   aws_access_key_id=aws_access_key_id,
   aws_secret_access_key=aws_secret_access_key,
   region_name=region_name
)

# S3バケット名とファイル名を指定
teacher_bucket_name = 'YOUR_TEACHER_BUCKET_NAME'
teacher_keys = ['path/to/teacher_image1.jpg', 'path/to/teacher_image2.jpg']  # 教師データのパス
video_bucket_name = 'YOUR_VIDEO_BUCKET_NAME'
video_key = 'path/to/your/video/file.mp4'

# 教師データをS3からダウンロードして読み込む
teacher_images = []
for key in teacher_keys:
   try:
       s3_object = s3.get_object(Bucket=teacher_bucket_name, Key=key)
       print(f"Retrieved {key} from S3")
       image_stream = BytesIO(s3_object['Body'].read())
       image = cv2.imdecode(np.frombuffer(image_stream.read(), np.uint8), cv2.IMREAD_COLOR)
       teacher_images.append(image)
   except NoCredentialsError:
       print("Credentials not available")
   except Exception as e:
       print(f"Error occurred while retrieving {key}: {e}")

# 動画をS3からダウンロードして一時ファイルに保存
try:
   s3_object = s3.get_object(Bucket=video_bucket_name, Key=video_key)
   print(f"Retrieved {video_key} from S3")
except NoCredentialsError:
   print("Credentials not available")
except Exception as e:
   print(f"Error occurred: {e}")

with tempfile.NamedTemporaryFile(delete=False) as temp_file:
   temp_file.write(s3_object['Body'].read())
   temp_video_path = temp_file.name

# 動画をビデオキャプチャとして読み込む
video_stream = cv2.VideoCapture(temp_video_path)

if not video_stream.isOpened():
   raise RuntimeError("Error: Could not open video stream")

# 特徴量の抽出とマッチング
def detect_and_compute_features(image):
   orb = cv2.ORB_create()
   keypoints, descriptors = orb.detectAndCompute(image, None)
   return keypoints, descriptors

# 教師データの特徴量を抽出
teacher_descriptors = []
for image in teacher_images:
   _, descriptors = detect_and_compute_features(image)
   teacher_descriptors.append(descriptors)

# フレームごとの特徴量を抽出して一致度を計算
frame_count = 0
similarity_scores = []

while video_stream.isOpened():
   ret, frame = video_stream.read()
   if ret:
       _, frame_descriptors = detect_and_compute_features(frame)
       if frame_descriptors is not None:
           bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
           total_matches = 0
           for descriptors in teacher_descriptors:
               matches = bf.match(frame_descriptors, descriptors)
               total_matches += len(matches)
           similarity_scores.append(total_matches)
       else:
           similarity_scores.append(0)
       frame_count += 1
   else:
       break

# 動画ストリームをリリースして一時ファイルを削除
video_stream.release()
os.remove(temp_video_path)

# グラフ化
plt.plot(similarity_scores)
plt.title('Similarity Scores Over Frames')
plt.xlabel('Frame Number')
plt.ylabel('Number of Matches')
plt.show()