Which GCP service to use - AutoML, Cloud Machine Learning APIs

[EXAMTOPIC] AutoML, Cloud ML API USE-CASES, 방법론 결정 포인트를 정리합니다.

Which GCP service to use - Situation & GCP service

• Cloud ML API : Pre-trained models, Ready to use, Efficient way
  • Don't have enough data
  • Don't need own labels
  • ML Task (Prediction task) is general enough to fit into the type of labels provided by the pretrained model.
  • Want to use labels from pre-trained APIs to train custom models with AutoML
• AutoML : Train high-quality Custom machine learning models with minimal effort and machine learning expertise.
  • Don't need to know specifics about the underlying models
  • Want to develop a quick initial model to use as a baseline ( which could end up being the production model )
• Custom ML Model
  • Not fit into any AutoML use-cases
  • Mixed input types e.g, images + tabular metadata
  • Want to control over model's architecture, framework, or exported model assets. (e.g, maybe your model needs to be built with TF.
  • Already have a baseline or heuristic and you want to see if you can improve upon it

Text

AutoML & NL API - Features

• BOTH AutoML & NL API

  • Classify documents(700 categories supported)
  • Entity extraction
  • Sentiment analysis (can be detected at entity level)
  • Analyzing text (Dependency parsing, Parsed label, POS tagging, Lemma, Morphology)
  • Integrated REST API : Natural Language is accessible via our REST API. Text can be uploaded in the request or integrated with Cloud Storage.

• AUTO ML : can customize classification categories, entities, and sentiment scores that are relevant to your application

Healthcare Natural Language

Healthcare Natural Language API	AutoML Entity Extraction for Healthcare
Specifically desinged for healthcare entity extraction - Extract information about medical concepts like diseases, medications, medical devices, procedures, and their clinically relevant attributes	Map medical concepts to standard medical vocabularies such as RxNorm, ICD-10, and MeSH

Use-cases

Which GCP service to use - Situation	GCP service
Integrated REST API - Natural Language is accessible via our REST API. Text can be uploaded in the request or integrated with Cloud Storage.	AutoML, Natural Language API
✔️ Syntax analysis - Extract tokens and sentences, identify parts of speech, and create dependency parse trees for each sentence.	Natural Language API
✔️ Entity analysis - Identify entities within documents—including receipts, invoices, and contracts—and label them by types such as date, person, and media.	Natural Language API
Custom entity extraction - Identify entities within documents and label them based on your own domain-specific keywords or phrases.	Auto ML
✔️ Sentiment analysis - Understand the overall opinion, feeling, or attitude sentiment expressed in a block of text.	Natural Language API
Custom sentiment analysis - Understand the overall opinion, feeling, or attitude expressed in a block of text tuned to your own domain-specific sentiment scores.	Auto ML
Content classification - Classify documents in 700+ predefined categories.	Natural Language API
Custom content classification - Create labels to customize models for unique use cases, using your own training data.	Auto ML
✔️ Multi-language - Analyze text in English, Spanish, Japanese, Chinese (simplified and traditional), French, German, Italian, Korean, Portuguese, and Russian.	AutoML, Natural Language API
Custom models - Train custom machine learning models with minimum effort and machine learning expertise.	Auto ML
Powered by Google’s AutoML models - Leverages Google state-of-the-art AutoML technology to produce high-quality models.	Auto ML
✔️ Spatial structure understanding - Use the structure and layout information in PDFs to improve custom entity extraction performance.	Auto ML
✔️ Large dataset support - Unlock complex use cases with support for 5,000 classification labels, 1 million documents, and 10 MB document size.	Auto ML

Vision

VS	Scenarios/Features	Which GCP Services to use
UI	Use APIs - Use REST and RPC APIs.	AutoML Vision, Vision API
UI	Use a graphical UI(user interface).	AutoML Vision
Labeling	Classify images using predefined labels. Pre-trained models leverage vast libraries of predefined labels.	Vision API
Labeling	Classify images using custom labels. Train models to classify images via labels you choose.	AutoML Vision
Labeling	Use Google’s data labeling service. Our team can help annotate your images, videos, and text.	AutoML Vision, Vision API
Deploy at the edge	Deploy machine learning models at the edge : Deploy low-latency, high-accuracy models optimized for edge devices.	AutoML Vision, Vision API (Integrate with ML Kit)
Features	Detect objects, where they are, and how many	AutoML Vision, Vision API
Features	Enable vision product search - Compare photos to images in your product catalog and return a ranked list of similar items.	Vision API
Features	Detect printed and handwritten text- Use OCR and automatically identify language.	Vision API
Features	Detect faces and facial attributes. (Face recognition not supported.)	Vision API
Features	Identify popular places and product logos - Automatically identify well-known landmarks and product logos.	Vision API
Features	Assign general image attributes - Detect general attributes and appropriate crop hints.	Vision API
Features	Detect web entities and pages - Find news events, logos, and similar images on the web.	Vision API
Features	Moderate content - Detect explicit content (adult, violent, etc.) within images.	Vision API
Features	Celebrity recognition - Identify celebrity faces in images (limited access, see documentation.)	Vision API

Video : AutoML Video Intelligence $vs.$ Video Intelligence API

Features

VS	GCP Products
Use REST and RPC APIs	AutoML, Video Intelligence API
Use a graphical UI	AutoML
Precise video analysis : Extract rich metadata at the video, shot, or frame level	AutoML, Video Intelligence API

Use-cases

USE CASE & Description	GCP Products
Content moderation : Identify when inappropriate content is being shown in a given video. You can instantly conduct content moderation across petabytes of data and more quickly and efficiently filter your content or user-generated content.	Video Intelligence API
Recommended content : Build a content recommendation engine with labels generated by ____ and a user’s viewing history and preferences. This will simplify content discovery for your users and guide them to the most relevant content that they want.	Video Intelligence API
Media archives : Create an indexed archive of your entire video library by using the metadata from ____. Ideal for mass media companies, it automatically analyzes content and make the results immediately accessible via the API.	Video Intelligence API
Contextual advertisements : You can identify appropriate locations in videos to insert ads that are contextually relevant to the video content. This can be done by matching the timeframe-specific labels of your video content with the content of your advertisements.	Video Intelligence API
Annotate video using predefined labels - Streaming video annotation : Pre-trained models leverage vast libraries of predefined labels.	Video Intelligence API
Annotate video using custom labels : Train models to classify video with custom labels of your choice.	AutoML
Stored video analysis, Streaming video analysis (beta), Shot change detection, Object detection and tracking	AutoML , Video Intelligence API
✔️ Text detection and extraction using OCR	Video Intelligence API
Explicit content detection	Video Intelligence API
Automated closed captioning and subtitles	Video Intelligence API
Logo recognition	Video Intelligence API
Celebrity recognition (limited access)	Video Intelligence API
Face detection (beta)	Video Intelligence API
Person detection with pose estimation (beta)	Video Intelligence API

Understanding Q

If you have an image classification task for identifying whether a car is present in a photo or not, which solution should you try first?

• ⭕ Try the Cloud Vision API first
• Try a custom model in TensorFlow first
• Try AutoML Vision first
• Try a custom model in BQML first

EXAMTOPIC Q 44

You work for a large technology company that wants to modernize their contact center. You have been asked to develop a solution to classify incoming calls by product so that requests can be more quickly routed to the correct support team. You have already transcribed the calls using the Speech-to-Text API. You want to minimize data preprocessing and development time. How should you build the model?

KEY : Minimize the preprocess & dev time to build the classifier using the result text data from speech-to-Text API

• Comparison on dev time : AI Platform vs. AutoML vs. ML API vs Custom Model
• Text classification by products

• A. Use the AI Platform Training built-in algorithms to create a custom model.

• ⭕ B. Use AutoML Natural Language to extract custom entities for classification.
  → AutoML classifies content in custom categories for your specific needs.

  • Using text : Natural Language
  • Custom entities/categories = products

• C. Use the Cloud Natural Language API to extract custom entities for classification.
  → Natural Language API reveals the structure and meaning of text with thousands of pretrained classifications.

• ❌ D. Build a custom model to identify the product keywords from the transcribed calls, and then run the keywords through a classification algorithm.
  → the most inefficient choice