翻訳は機械翻訳により提供されています。提供された翻訳内容と英語版の間で齟齬、不一致または矛盾がある場合、英語版が優先します。
例: 保存されたビデオ内のセグメントの検出
以下の手順は、Amazon S3 バケットに保存されているビデオ内のテクニカルキューセグメントとショット検出セグメントを検出する方法を示しています。この手順では、Amazon Rekognition Video が持つ検出精度に対する信頼度に基づいて、検出されたセグメントをフィルタリングする方法も示します。
この例では、Amazon Simple Queue Service のキューを使用してビデオ分析リクエストの完了ステータスを取得する Java または Python を使用して Amazon S3 バケットに保存されている動画を分析する (SDK) のコードを拡張します。
Amazon S3 バケットに保存されたビデオ内のセグメントを検出するには (SDK)
-
「Java または Python を使用して Amazon S3 バケットに保存されている動画を分析する (SDK)」を実行します。
-
ステップ 1 で使用したコードに以下を追加します。
- Java
-
-
次のインポートを追加します。
import com.amazonaws.services.rekognition.model.GetSegmentDetectionRequest; import com.amazonaws.services.rekognition.model.GetSegmentDetectionResult; import com.amazonaws.services.rekognition.model.SegmentDetection; import com.amazonaws.services.rekognition.model.SegmentType; import com.amazonaws.services.rekognition.model.SegmentTypeInfo; import com.amazonaws.services.rekognition.model.ShotSegment; import com.amazonaws.services.rekognition.model.StartSegmentDetectionFilters; import com.amazonaws.services.rekognition.model.StartSegmentDetectionRequest; import com.amazonaws.services.rekognition.model.StartSegmentDetectionResult; import com.amazonaws.services.rekognition.model.StartShotDetectionFilter; import com.amazonaws.services.rekognition.model.StartTechnicalCueDetectionFilter; import com.amazonaws.services.rekognition.model.TechnicalCueSegment; import com.amazonaws.services.rekognition.model.AudioMetadata;
-
次のコードを
VideoDetect
クラスに追加します。//Copyright 2020 Amazon.com, Inc. or its affiliates. All Rights Reserved. //PDX-License-Identifier: MIT-0 (For details, see https://github.com/awsdocs/amazon-rekognition-developer-guide/blob/master/LICENSE-SAMPLECODE.) private static void StartSegmentDetection(String bucket, String video) throws Exception{ NotificationChannel channel= new NotificationChannel() .withSNSTopicArn(snsTopicArn) .withRoleArn(roleArn); float minTechnicalCueConfidence = 80F; float minShotConfidence = 80F; StartSegmentDetectionRequest req = new StartSegmentDetectionRequest() .withVideo(new Video() .withS3Object(new S3Object() .withBucket(bucket) .withName(video))) .withSegmentTypes("TECHNICAL_CUE" , "SHOT") .withFilters(new StartSegmentDetectionFilters() .withTechnicalCueFilter(new StartTechnicalCueDetectionFilter() .withMinSegmentConfidence(minTechnicalCueConfidence)) .withShotFilter(new StartShotDetectionFilter() .withMinSegmentConfidence(minShotConfidence))) .withJobTag("DetectingVideoSegments") .withNotificationChannel(channel); StartSegmentDetectionResult startLabelDetectionResult = rek.startSegmentDetection(req); startJobId=startLabelDetectionResult.getJobId(); } private static void GetSegmentDetectionResults() throws Exception{ int maxResults=10; String paginationToken=null; GetSegmentDetectionResult segmentDetectionResult=null; Boolean firstTime=true; do { if (segmentDetectionResult !=null){ paginationToken = segmentDetectionResult.getNextToken(); } GetSegmentDetectionRequest segmentDetectionRequest= new GetSegmentDetectionRequest() .withJobId(startJobId) .withMaxResults(maxResults) .withNextToken(paginationToken); segmentDetectionResult = rek.getSegmentDetection(segmentDetectionRequest); if(firstTime) { System.out.println("\nStatus\n------"); System.out.println(segmentDetectionResult.getJobStatus()); System.out.println("\nRequested features\n------------------"); for (SegmentTypeInfo requestedFeatures : segmentDetectionResult.getSelectedSegmentTypes()) { System.out.println(requestedFeatures.getType()); } int count=1; List<VideoMetadata> videoMetaDataList = segmentDetectionResult.getVideoMetadata(); System.out.println("\nVideo Streams\n-------------"); for (VideoMetadata videoMetaData: videoMetaDataList) { System.out.println("Stream: " + count++); System.out.println("\tFormat: " + videoMetaData.getFormat()); System.out.println("\tCodec: " + videoMetaData.getCodec()); System.out.println("\tDuration: " + videoMetaData.getDurationMillis()); System.out.println("\tFrameRate: " + videoMetaData.getFrameRate()); } List<AudioMetadata> audioMetaDataList = segmentDetectionResult.getAudioMetadata(); System.out.println("\nAudio streams\n-------------"); count=1; for (AudioMetadata audioMetaData: audioMetaDataList) { System.out.println("Stream: " + count++); System.out.println("\tSample Rate: " + audioMetaData.getSampleRate()); System.out.println("\tCodec: " + audioMetaData.getCodec()); System.out.println("\tDuration: " + audioMetaData.getDurationMillis()); System.out.println("\tNumber of Channels: " + audioMetaData.getNumberOfChannels()); } System.out.println("\nSegments\n--------"); firstTime=false; } //Show segment information List<SegmentDetection> detectedSegments= segmentDetectionResult.getSegments(); for (SegmentDetection detectedSegment: detectedSegments) { if (detectedSegment.getType().contains(SegmentType.TECHNICAL_CUE.toString())) { System.out.println("Technical Cue"); TechnicalCueSegment segmentCue=detectedSegment.getTechnicalCueSegment(); System.out.println("\tType: " + segmentCue.getType()); System.out.println("\tConfidence: " + segmentCue.getConfidence().toString()); } if (detectedSegment.getType().contains(SegmentType.SHOT.toString())) { System.out.println("Shot"); ShotSegment segmentShot=detectedSegment.getShotSegment(); System.out.println("\tIndex " + segmentShot.getIndex()); System.out.println("\tConfidence: " + segmentShot.getConfidence().toString()); } long seconds=detectedSegment.getDurationMillis(); System.out.println("\tDuration : " + Long.toString(seconds) + " milliseconds"); System.out.println("\tStart time code: " + detectedSegment.getStartTimecodeSMPTE()); System.out.println("\tEnd time code: " + detectedSegment.getEndTimecodeSMPTE()); System.out.println("\tDuration time code: " + detectedSegment.getDurationSMPTE()); System.out.println(); } } while (segmentDetectionResult !=null && segmentDetectionResult.getNextToken() != null); }
-
関数
main
で、以下の行を置き換えます。StartLabelDetection(bucket, video); if (GetSQSMessageSuccess()==true) GetLabelDetectionResults();
を:
StartSegmentDetection(bucket, video); if (GetSQSMessageSuccess()==true) GetSegmentDetectionResults();
-
- Java V2
-
//snippet-start:[rekognition.java2.recognize_video_text.import] import software.amazon.awssdk.auth.credentials.ProfileCredentialsProvider; import software.amazon.awssdk.regions.Region; import software.amazon.awssdk.services.rekognition.RekognitionClient; import software.amazon.awssdk.services.rekognition.model.S3Object; import software.amazon.awssdk.services.rekognition.model.NotificationChannel; import software.amazon.awssdk.services.rekognition.model.Video; import software.amazon.awssdk.services.rekognition.model.StartTextDetectionRequest; import software.amazon.awssdk.services.rekognition.model.StartTextDetectionResponse; import software.amazon.awssdk.services.rekognition.model.RekognitionException; import software.amazon.awssdk.services.rekognition.model.GetTextDetectionResponse; import software.amazon.awssdk.services.rekognition.model.GetTextDetectionRequest; import software.amazon.awssdk.services.rekognition.model.VideoMetadata; import software.amazon.awssdk.services.rekognition.model.TextDetectionResult; import java.util.List; //snippet-end:[rekognition.java2.recognize_video_text.import] /** * Before running this Java V2 code example, set up your development environment, including your credentials. * * For more information, see the following documentation topic: * * https://docs.aws.amazon.com/sdk-for-java/latest/developer-guide/get-started.html */ public class DetectVideoSegments { private static String startJobId =""; public static void main(String[] args) { final String usage = "\n" + "Usage: " + " <bucket> <video> <topicArn> <roleArn>\n\n" + "Where:\n" + " bucket - The name of the bucket in which the video is located (for example, (for example, myBucket). \n\n"+ " video - The name of video (for example, people.mp4). \n\n" + " topicArn - The ARN of the Amazon Simple Notification Service (Amazon SNS) topic. \n\n" + " roleArn - The ARN of the AWS Identity and Access Management (IAM) role to use. \n\n" ; if (args.length != 4) { System.out.println(usage); System.exit(1); } String bucket = args[0]; String video = args[1]; String topicArn = args[2]; String roleArn = args[3]; Region region = Region.US_WEST_2; RekognitionClient rekClient = RekognitionClient.builder() .region(region) .credentialsProvider(ProfileCredentialsProvider.create("profile-name")) .build(); NotificationChannel channel = NotificationChannel.builder() .snsTopicArn(topicArn) .roleArn(roleArn) .build(); startTextLabels(rekClient, channel, bucket, video); GetTextResults(rekClient); System.out.println("This example is done!"); rekClient.close(); } // snippet-start:[rekognition.java2.recognize_video_text.main] public static void startTextLabels(RekognitionClient rekClient, NotificationChannel channel, String bucket, String video) { try { S3Object s3Obj = S3Object.builder() .bucket(bucket) .name(video) .build(); Video vidOb = Video.builder() .s3Object(s3Obj) .build(); StartTextDetectionRequest labelDetectionRequest = StartTextDetectionRequest.builder() .jobTag("DetectingLabels") .notificationChannel(channel) .video(vidOb) .build(); StartTextDetectionResponse labelDetectionResponse = rekClient.startTextDetection(labelDetectionRequest); startJobId = labelDetectionResponse.jobId(); } catch (RekognitionException e) { System.out.println(e.getMessage()); System.exit(1); } } public static void GetTextResults(RekognitionClient rekClient) { try { String paginationToken=null; GetTextDetectionResponse textDetectionResponse=null; boolean finished = false; String status; int yy=0 ; do{ if (textDetectionResponse !=null) paginationToken = textDetectionResponse.nextToken(); GetTextDetectionRequest recognitionRequest = GetTextDetectionRequest.builder() .jobId(startJobId) .nextToken(paginationToken) .maxResults(10) .build(); // Wait until the job succeeds. while (!finished) { textDetectionResponse = rekClient.getTextDetection(recognitionRequest); status = textDetectionResponse.jobStatusAsString(); if (status.compareTo("SUCCEEDED") == 0) finished = true; else { System.out.println(yy + " status is: " + status); Thread.sleep(1000); } yy++; } finished = false; // Proceed when the job is done - otherwise VideoMetadata is null. VideoMetadata videoMetaData=textDetectionResponse.videoMetadata(); System.out.println("Format: " + videoMetaData.format()); System.out.println("Codec: " + videoMetaData.codec()); System.out.println("Duration: " + videoMetaData.durationMillis()); System.out.println("FrameRate: " + videoMetaData.frameRate()); System.out.println("Job"); List<TextDetectionResult> labels= textDetectionResponse.textDetections(); for (TextDetectionResult detectedText: labels) { System.out.println("Confidence: " + detectedText.textDetection().confidence().toString()); System.out.println("Id : " + detectedText.textDetection().id()); System.out.println("Parent Id: " + detectedText.textDetection().parentId()); System.out.println("Type: " + detectedText.textDetection().type()); System.out.println("Text: " + detectedText.textDetection().detectedText()); System.out.println(); } } while (textDetectionResponse !=null && textDetectionResponse.nextToken() != null); } catch(RekognitionException | InterruptedException e) { System.out.println(e.getMessage()); System.exit(1); } } // snippet-end:[rekognition.java2.recognize_video_text.main] }
- Python
-
-
ステップ 1 で作成したクラス
VideoDetect
に次のコードを追加します。# Copyright 2020 Amazon.com, Inc. or its affiliates. All Rights Reserved. # PDX-License-Identifier: MIT-0 (For details, see https://github.com/awsdocs/amazon-rekognition-developer-guide/blob/master/LICENSE-SAMPLECODE.) def StartSegmentDetection(self): min_Technical_Cue_Confidence = 80.0 min_Shot_Confidence = 80.0 max_pixel_threshold = 0.1 min_coverage_percentage = 60 response = self.rek.start_segment_detection( Video={"S3Object": {"Bucket": self.bucket, "Name": self.video}}, NotificationChannel={ "RoleArn": self.roleArn, "SNSTopicArn": self.snsTopicArn, }, SegmentTypes=["TECHNICAL_CUE", "SHOT"], Filters={ "TechnicalCueFilter": { "BlackFrame": { "MaxPixelThreshold": max_pixel_threshold, "MinCoveragePercentage": min_coverage_percentage, }, "MinSegmentConfidence": min_Technical_Cue_Confidence, }, "ShotFilter": {"MinSegmentConfidence": min_Shot_Confidence}, } ) self.startJobId = response["JobId"] print(f"Start Job Id: {self.startJobId}") def GetSegmentDetectionResults(self): maxResults = 10 paginationToken = "" finished = False firstTime = True while finished == False: response = self.rek.get_segment_detection( JobId=self.startJobId, MaxResults=maxResults, NextToken=paginationToken ) if firstTime == True: print(f"Status\n------\n{response['JobStatus']}") print("\nRequested Types\n---------------") for selectedSegmentType in response['SelectedSegmentTypes']: print(f"\tType: {selectedSegmentType['Type']}") print(f"\t\tModel Version: {selectedSegmentType['ModelVersion']}") print() print("\nAudio metadata\n--------------") for audioMetadata in response['AudioMetadata']: print(f"\tCodec: {audioMetadata['Codec']}") print(f"\tDuration: {audioMetadata['DurationMillis']}") print(f"\tNumber of Channels: {audioMetadata['NumberOfChannels']}") print(f"\tSample rate: {audioMetadata['SampleRate']}") print() print("\nVideo metadata\n--------------") for videoMetadata in response["VideoMetadata"]: print(f"\tCodec: {videoMetadata['Codec']}") print(f"\tColor Range: {videoMetadata['ColorRange']}") print(f"\tDuration: {videoMetadata['DurationMillis']}") print(f"\tFormat: {videoMetadata['Format']}") print(f"\tFrame rate: {videoMetadata['FrameRate']}") print("\nSegments\n--------") firstTime = False for segment in response['Segments']: if segment["Type"] == "TECHNICAL_CUE": print("Technical Cue") print(f"\tConfidence: {segment['TechnicalCueSegment']['Confidence']}") print(f"\tType: {segment['TechnicalCueSegment']['Type']}") if segment["Type"] == "SHOT": print("Shot") print(f"\tConfidence: {segment['ShotSegment']['Confidence']}") print(f"\tIndex: " + str(segment["ShotSegment"]["Index"])) print(f"\tDuration (milliseconds): {segment['DurationMillis']}") print(f"\tStart Timestamp (milliseconds): {segment['StartTimestampMillis']}") print(f"\tEnd Timestamp (milliseconds): {segment['EndTimestampMillis']}") print(f"\tStart timecode: {segment['StartTimecodeSMPTE']}") print(f"\tEnd timecode: {segment['EndTimecodeSMPTE']}") print(f"\tDuration timecode: {segment['DurationSMPTE']}") print(f"\tStart frame number {segment['StartFrameNumber']}") print(f"\tEnd frame number: {segment['EndFrameNumber']}") print(f"\tDuration frames: {segment['DurationFrames']}") print() if "NextToken" in response: paginationToken = response["NextToken"] else: finished = True
-
関数
main
で、以下の行を置き換えます。analyzer.StartLabelDetection() if analyzer.GetSQSMessageSuccess()==True: analyzer.GetLabelDetectionResults()
を:
analyzer.StartSegmentDetection() if analyzer.GetSQSMessageSuccess()==True: analyzer.GetSegmentDetectionResults()
-
注記
Java または Python を使用して Amazon S3 バケットに保存されている動画を分析する (SDK) 以外のビデオ例をすでに実行している場合、置き換えるコードは異なる可能性があります。
-
コードを実行します。入力ビデオで検出されたセグメントに関する情報が表示されます。
Amazon Rekognition セグメント API を利用する
顔のライブネスの検出