Quickly creating initial ML models with AutoML Vision
Short answer
B Train an object detection model in AutoML by using the annotated image data.
Question
Key Takeaways
AutoML is optimized for speed and ease of use when data is already labeled, whereas custom training and pipelines add unnecessary complexity for initial prototypes.
Learn why Vertex AI AutoML is the best choice for rapidly building initial object detection models with pre-labeled image data in Google Cloud.
Common mistake: Candidates often choose Custom Training (D) or Pipelines (C) thinking they offer more control, but these options ignore the 'quickly create' constraint and require significantly more engineering effort.
Expert Analysis
The scenario requires creating an initial model quickly for a specific computer vision task (object detection) using annotated images (bounding boxes and labels). The team has sufficient budget and existing labeled data.
Option B is the correct choice because Vertex AI AutoML Vision Object Detection is specifically designed for this use case. It allows users to train high-quality models with minimal code or ML expertise by simply uploading the dataset. AutoML automatically handles architecture selection, hyperparameter tuning, and training, which aligns perfectly with the requirement to "quickly create an initial model." The comments correctly highlight that AutoML is ideal when data is already labeled and speed of development is prioritized over custom architectural control.
Option A involves fine-tuning a pre-trained model from TensorFlow Hub. While feasible, this requires writing custom training code in Vertex AI Workbench, which is slower and more complex than AutoML. Option D suggests Vertex AI Custom Training, which also requires significant coding effort to define the model architecture and training loop, contradicting the goal of speed. Option C involves setting up a Vertex AI Pipeline. This adds unnecessary orchestration overhead for a simple proof-of-concept; pipelines are better suited for complex, repeatable production workflows rather than rapid initial prototyping.
Therefore, leveraging the managed AutoML service is the most efficient path to an initial model given the constraints.
Exam Strategy
When exam questions emphasize 'quickly,' 'initial prototype,' or 'minimal code' with pre-labeled data, prioritize managed services like AutoML over custom training solutions. Evaluate if the added complexity of pipelines or custom notebooks is justified by the scenario's constraints; usually, it is not for early-stage proofs-of-concept.
Frequently Asked Questions
Why not use a pre-trained model from TensorFlow Hub?
Is AutoML suitable for production workloads?
Community Votes
100% of anonymous learners picked answer B. Votes are pick records left by other test-takers — they are not the verified answer.
Discussion 8 contributions
Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.
From public community threads 8
-
pikachu007 👍 6 Answered BSpeed: AutoML excels in creating high-quality models with minimal code and setup, significantly accelerating model development. Ease of use: It provides a user-friendly interface and automates many aspects of model training, making it accessible even for those without extensive ML expertise. Automatic optimization: AutoML automatically handles hyperparameter tuning, feature engineering, and architecture selection, reducing manual effort and expertise required. Custom object detection: It supports custom object detection tasks, directly addressing the need to identify damaged parts in images.
-
VinaoSilva 👍 3 Answered Bquickly create an initial model = automl
-
edoo 👍 2 Answered BBy doing B we are doing D. I suppose B in more specific about the model and thus "more" correct? Thoughts?
-
pinimichele01 👍 2 Answered Bwent with B
-
taylz876 👍 1 Answered BThe team already has annotated images, with a bounding box and a part name on each damaged part. That is exactly the training data format AutoML image object detection expects. You need an initial model quickly and have the budget, and AutoML trains it for you with no code, handling the architecture, training and tuning. Today this is AutoML image object detection in Agent Platform (formerly Vertex AI). Why the others are wrong • A: Downloading a pre-trained model from TensorFlow Hub and fine-tuning it in Workbench means writing and managing the code yourself. It works, but it's slower than AutoML. • C: Building a pipeline to run an AutoML training job adds extra work for a proof of concept. Pipelines are for repeatable workflows, and you just need a first model. • D: Custom training means writing your own training code, which is the slowest route here.
Show 3 more comments
-
louisaok 👍 1 Answered B>>" You have been given a sufficient budget to train models on Google Cloud" it is rare to see a company give enough money to run a mission-critical project.
-
ddogg 👍 1 Answered BB makes the most sense, data is already labelled and a pretrained model may not fit for this specific case
-
Foxy2021 👍 1My vote is B
Reader comments & corrections
No reader corrections yet — spotted an error or have a note? Share it below.
Related Analysis
Practice All PMLE Questions
Access 65 questions with complete answers and detailed explanations.
View Full PMLE Practice Test →