Google Professional Machine Learning Engineer Question 338
Single answerGoogle Cloud PlatformYou are designing a machine learning system for a client that requires low latency for real-time object detection on video streams from multiple cameras. The model must process data directly on the devices located at the client's facilities, where internet connectivity is limited. Which hardware option should you choose to meet these requirements?
- A
Cloud-based GPUs with high computational power
- B
Edge TPUs deployed on local devices
- C
Standard CPUs with multi-threading capabilities
- D
Google Cloud TPUs accessed via a cloud-based infrastructure
Show answer and explanation
Correct answer: B
Explanation
In this scenario, the primary requirements include low latency, real-time processing, and on-device computation due to limited internet connectivity. Edge TPUs are specifically designed for such use cases, offering optimized performance for machine learning tasks directly on edge devices. Other options, such as cloud-based GPUs or TPUs, rely on internet connectivity and are unsuitable for this localized, low-latency scenario.
- A. Incorrect.
Cloud-based GPUs offer high computational power, but they are not suitable for scenarios with limited internet connectivity and the need for on-device, low-latency processing.
- B. Correct.
Edge TPUs are designed specifically for on-device machine learning tasks with low latency, making them the most appropriate choice for this scenario.
- C. Incorrect.
Standard CPUs, while versatile, are not optimized for real-time object detection and would likely fail to meet the low-latency requirement in this context.
- D. Incorrect.
Google Cloud TPUs are powerful for large-scale training and inference tasks in the cloud, but they are not suitable for on-device processing with limited connectivity.