動画の分類により、物体、場所、活動、動物の種類、商品などを識別できます。
AutoML 動画の使用
始める前に
AutoML モデルの作成に関するバックグラウンドについては、Vertex AI の初心者向けガイドをご覧ください。AutoML モデルを作成する方法については、Vertex AI ドキュメントの「ML モデルの開発と使用」にある動画データをご覧ください。
AutoML モデルの使用
次のコードサンプルは、ストリーミング クライアント ライブラリを使用して動画分類を行うために AutoML モデルを使用する方法を示しています。
Java
Video Intelligence への認証を行うには、アプリケーションのデフォルト認証情報を設定します。 詳細については、ローカル開発環境の認証を設定するをご覧ください。
import com.google.api.gax.rpc.BidiStream;
import com.google.cloud.videointelligence.v1p3beta1.LabelAnnotation;
import com.google.cloud.videointelligence.v1p3beta1.LabelFrame;
import com.google.cloud.videointelligence.v1p3beta1.StreamingAnnotateVideoRequest;
import com.google.cloud.videointelligence.v1p3beta1.StreamingAnnotateVideoResponse;
import com.google.cloud.videointelligence.v1p3beta1.StreamingAutomlClassificationConfig;
import com.google.cloud.videointelligence.v1p3beta1.StreamingFeature;
import com.google.cloud.videointelligence.v1p3beta1.StreamingVideoAnnotationResults;
import com.google.cloud.videointelligence.v1p3beta1.StreamingVideoConfig;
import com.google.cloud.videointelligence.v1p3beta1.StreamingVideoIntelligenceServiceClient;
import com.google.protobuf.ByteString;
import io.grpc.StatusRuntimeException;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.util.Arrays;
import java.util.concurrent.TimeoutException;
class StreamingAutoMlClassification {
// Perform streaming video classification with an AutoML Model
static void streamingAutoMlClassification(String filePath, String projectId, String modelId)
throws TimeoutException, StatusRuntimeException, IOException {
// String filePath = "path_to_your_video_file";
// String projectId = "YOUR_GCP_PROJECT_ID";
// String modelId = "YOUR_AUTO_ML_CLASSIFICATION_MODEL_ID";
try (StreamingVideoIntelligenceServiceClient client =
StreamingVideoIntelligenceServiceClient.create()) {
Path path = Paths.get(filePath);
byte[] data = Files.readAllBytes(path);
// Set the chunk size to 5MB (recommended less than 10MB).
int chunkSize = 5 * 1024 * 1024;
int numChunks = (int) Math.ceil((double) data.length / chunkSize);
String modelPath =
String.format("projects/%s/locations/us-central1/models/%s", projectId, modelId);
System.out.println(modelPath);
StreamingAutomlClassificationConfig streamingAutomlClassificationConfig =
StreamingAutomlClassificationConfig.newBuilder().setModelName(modelPath).build();
StreamingVideoConfig streamingVideoConfig =
StreamingVideoConfig.newBuilder()
.setFeature(StreamingFeature.STREAMING_AUTOML_CLASSIFICATION)
.setAutomlClassificationConfig(streamingAutomlClassificationConfig)
.build();
BidiStream<StreamingAnnotateVideoRequest, StreamingAnnotateVideoResponse> call =
client.streamingAnnotateVideoCallable().call();
// The first request must **only** contain the audio configuration:
call.send(
StreamingAnnotateVideoRequest.newBuilder().setVideoConfig(streamingVideoConfig).build());
// Subsequent requests must **only** contain the audio data.
// Send the requests in chunks
for (int i = 0; i < numChunks; i++) {
call.send(
StreamingAnnotateVideoRequest.newBuilder()
.setInputContent(
ByteString.copyFrom(
Arrays.copyOfRange(data, i * chunkSize, i * chunkSize + chunkSize)))
.build());
}
// Tell the service you are done sending data
call.closeSend();
for (StreamingAnnotateVideoResponse response : call) {
if (response.hasError()) {
System.out.println(response.getError().getMessage());
break;
}
StreamingVideoAnnotationResults annotationResults = response.getAnnotationResults();
for (LabelAnnotation annotation : annotationResults.getLabelAnnotationsList()) {
String entity = annotation.getEntity().getDescription();
// There is only one frame per annotation
LabelFrame labelFrame = annotation.getFrames(0);
double offset =
labelFrame.getTimeOffset().getSeconds() + labelFrame.getTimeOffset().getNanos() / 1e9;
float confidence = labelFrame.getConfidence();
System.out.format("At %fs segment: %s (%f)\n", offset, entity, confidence);
}
}
System.out.println("Video streamed successfully.");
}
}
}Node.js
Video Intelligence への認証を行うには、アプリケーションのデフォルト認証情報を設定します。 詳細については、ローカル開発環境の認証を設定するをご覧ください。
/**
* TODO(developer): Uncomment these variables before running the sample.
*/
// const path = 'Local file to analyze, e.g. ./my-file.mp4';
// const modelId = 'autoMl model'
// const projectId = 'Your GCP Project'
const {StreamingVideoIntelligenceServiceClient} =
require('@google-cloud/video-intelligence').v1p3beta1;
const fs = require('fs');
// Instantiates a client
const client = new StreamingVideoIntelligenceServiceClient();
// Streaming configuration
const modelPath = `projects/${projectId}/locations/us-central1/models/${modelId}`;
const configRequest = {
videoConfig: {
feature: 'STREAMING_AUTOML_CLASSIFICATION',
automlClassificationConfig: {
modelName: modelPath,
},
},
};
const readStream = fs.createReadStream(path, {
highWaterMark: 5 * 1024 * 1024, //chunk size set to 5MB (recommended less than 10MB)
encoding: 'base64',
});
//Load file content
// Note: Input videos must have supported video codecs. See
// https://cloud.google.com/video-intelligence/docs/streaming/streaming#supported_video_codecs
// for more details.
const chunks = [];
readStream
.on('data', chunk => {
const request = {
inputContent: chunk.toString(),
};
chunks.push(request);
})
.on('close', () => {
// configRequest should be the first in the stream of requests
stream.write(configRequest);
for (let i = 0; i < chunks.length; i++) {
stream.write(chunks[i]);
}
stream.end();
});
const stream = client
.streamingAnnotateVideo()
.on('data', response => {
//Gets annotations for video
const annotations = response.annotationResults;
const labels = annotations.labelAnnotations;
labels.forEach(label => {
console.log(
`Label ${label.entity.description} occurs at: ${
label.frames[0].timeOffset.seconds || 0
}` + `.${(label.frames[0].timeOffset.nanos / 1e6).toFixed(0)}s`
);
console.log(` Confidence: ${label.frames[0].confidence}`);
});
})
.on('error', response => {
console.error(response);
});Python
Video Intelligence への認証を行うには、アプリケーションのデフォルト認証情報を設定します。 詳細については、ローカル開発環境の認証を設定するをご覧ください。
import io
from google.cloud import videointelligence_v1p3beta1 as videointelligence
# path = 'path_to_file'
# project_id = 'gcp_project_id'
# model_id = 'automl_classification_model_id'
client = videointelligence.StreamingVideoIntelligenceServiceClient()
model_path = "projects/{}/locations/us-central1/models/{}".format(
project_id, model_id
)
# Here we use classification as an example.
automl_config = videointelligence.StreamingAutomlClassificationConfig(
model_name=model_path
)
video_config = videointelligence.StreamingVideoConfig(
feature=videointelligence.StreamingFeature.STREAMING_AUTOML_CLASSIFICATION,
automl_classification_config=automl_config,
)
# config_request should be the first in the stream of requests.
config_request = videointelligence.StreamingAnnotateVideoRequest(
video_config=video_config
)
# Set the chunk size to 5MB (recommended less than 10MB).
chunk_size = 5 * 1024 * 1024
# Load file content.
# Note: Input videos must have supported video codecs. See
# https://cloud.google.com/video-intelligence/docs/streaming/streaming#supported_video_codecs
# for more details.
stream = []
with io.open(path, "rb") as video_file:
while True:
data = video_file.read(chunk_size)
if not data:
break
stream.append(data)
def stream_generator():
yield config_request
for chunk in stream:
yield videointelligence.StreamingAnnotateVideoRequest(input_content=chunk)
requests = stream_generator()
# streaming_annotate_video returns a generator.
# The default timeout is about 300 seconds.
# To process longer videos it should be set to
# larger than the length (in seconds) of the stream.
responses = client.streaming_annotate_video(requests, timeout=600)
for response in responses:
# Check for errors.
if response.error.message:
print(response.error.message)
break
for label in response.annotation_results.label_annotations:
for frame in label.frames:
print(
"At {:3d}s segment, {:5.1%} {}".format(
frame.time_offset.seconds,
frame.confidence,
label.entity.entity_id,
)
)