Unleashing Google's Vision AI for Personalized Impact (By: Rabbia Arshad) - DevFest Lahore 2023

Workshop by Rabbia Arshad (https://www.linkedin.com/in/rabbiaarshad/) at DevFest Lahore 2023 by GDG Lahore.

Transcript

1. Unleashing Google’s Vision AI for Personalised Impact Rabbia Arshad Machine Learning Engineer @ SmartEnds
2. Agenda ⬡ Basics of Computer Vision ⬡ Google Vision AI potential ⬡ Challenges in Computer Vision ⬡ Model Training Cycle ⬡ Use Case ⬡ Example 2
3. Computer Vision Computer vision is a field of computer science that focuses on enabling computers to identify and understand objects and people in images and videos. It replicate both the way humans see, and the way humans make sense of what they see. 3
4. “In the realm of technology, computer vision is the visual poetry that transforms pixels into possibilities. 4
5. Google Vision AI Fully managed development environment Vertex AI Vision to create your own computer vision applications or derive insights from images and videos
6. Features ⬡ Spin up new video and image analytics applications in minutes ⬡ Train machine learning models that classify images using AutoML or custom models ⬡ Easily integrate with BigQuery, Cloud Functions, and your cameras to enable end to end journey 6
7. ⬡ Large Data Availability and Quality Challenges in Training Computer Vision Model ⬡ High Computation System to Trained Model 7 ⬡ Follow Evolving Standards and Frameworks ⬡ Time Constraint - Require a lot of time in research building stuff
8. Benefits Faster time to value with reduced complexity Easily build, deploy and manage computer vision applications. Solve for a variety of uses and expertise levels Plug and play analytics via APIs or the ability to use custom ML models or an end to end development environment. Assured quality from by Following Standards Following Standard and Framework, Data Protection Policies, Update and Management via new techniques. 8
9. ⬡ Label Classification ∙ Single Label ∙ Multi Label Services ⬡ Image Text Extraction ∙ OCR 9 ⬡ Object Detection ∙ Localization ⬡ Image Segmentation
11. 11 Vision API Vision API offers powerful pre-trained machine learning models through REST and RPC APIs. 3 Auto ML Automate the training of your own custom machine learning models. Simply upload images and train custom image and video models with AutoML's easy-to-use graphical interface; optimize your models for accuracy, latency, and size 2 Vertex Vision AI Vertex AI Vision is a fully managed end to end application development environment that lets you easily build, deploy and manage computer vision applications for your unique business needs. 1 Solutions
12. Model Cycle 12 1 3 5 6 4 2 Prepare Data Data collection, storage Train Model Tune Hyperparameters Get Prediction Test on test set, check accuracy, precision Create Dataset Add Data Annotation Labeling Evaluate Model Test on Validation Data Deployment & Maintenance
13. Use Case - Vision Product Search 13
14. Example ⬡ Single Label ⬡ Multi Classification ⬡ 5 Classes ∙ Roses ∙ Daisy ∙ Tulips ∙ Dandelion ∙ Roses ∙ Sunflowers 14
15. Confusion Matrix 15
16. Result 16
17. 17 Thanks ! Any questions? You can find me at: Rabbia Arshad linkedin.com/in/rabbiaarshad rabbiaarshad09@gmail.com