Google Professional Machine Learning Engineer
Get started today
Ultimate access to all questions.
You are a Machine Learning Engineer at a startup that wants to deploy an object detection model to identify the presence and location of its logo in images. The startup has provided you with a large, unlabeled dataset of images, some containing the logo and others not. The startup is under budget constraints and requires a solution that balances cost, efficiency, and accuracy. Additionally, the solution must be scalable to accommodate future increases in dataset size. Considering these constraints, what is the most efficient approach to label the data, train, and deploy the model? Choose the best option.
You are a Machine Learning Engineer at a startup that wants to deploy an object detection model to identify the presence and location of its logo in images. The startup has provided you with a large, unlabeled dataset of images, some containing the logo and others not. The startup is under budget constraints and requires a solution that balances cost, efficiency, and accuracy. Additionally, the solution must be scalable to accommodate future increases in dataset size. Considering these constraints, what is the most efficient approach to label the data, train, and deploy the model? Choose the best option.
Real Exam
Explanation:
Correct Answer: A
Why?
- Data Labelling Service: Provides an efficient and scalable solution for labeling large datasets, reducing manual effort and potential errors.
- AutoML Object Detection: Offers a fully managed solution that is cost-effective and suitable for startups with limited ML expertise, ensuring quick deployment and scalability.
- Efficiency and Scalability: These services are designed to handle large datasets and can scale with the startup's growth, aligning with the budget and future requirements.
Other Options Considered:
- Vision API: While it can detect objects, it may not offer the customization needed for specific logo detection tasks, potentially affecting accuracy.
- Manual Labeling: Not scalable or efficient for large datasets, and prone to human error.
- Real-Time Object Detection: Unnecessary for the startup's current needs, adding unnecessary complexity and cost.