How to Label Employee-Only Image Data With Maximum Accuracy?
An ML engineer needs to train a supervised deep learning model. The available dataset is a large number of unlabeled images that only employees should access. The ML engineer needs to implement a solution that labels the dataset with the highest possible accuracy. Which combination of steps should the ML engineer take to meet these requirements? (Choose two.)
Community Insight
The question tests the pairing of a managed labeling service with the correct workforce type for confidential data, and the trap is picking a public crowdsourcing option or assuming a model can label itself in a supervised setting.
Building a supervised image model requires accurate ground-truth labels, and for confidential employee-only images the labeling must happen inside a controlled private workforce. This page establishes that Amazon SageMaker Ground Truth annotation jobs staffed by a private team (C and D) satisfy both the accuracy and the access-restriction requirements.
Choosing Amazon Mechanical Turk (E) because it is cheap and crowdsourced, ignoring that the images are restricted to employees — a public workforce would break the access requirement even if labeling throughput were higher.
Community Discussion (6 comments)
Comments & Corrections
No comments yet — spotted an error or have a note? Share it below.
Expert Analysis
Why the Answer Is Correct
Supervised deep learning needs human-provided labels, and the requirement for "highest possible accuracy" on images that "only employees should access" maps directly to Amazon SageMaker Ground Truth plus a private workforce. Option C creates the annotation job with the labeling task and its requirements defined, which is the mechanism Ground Truth uses to collect high-quality labels. Option D supplies the workers: a private workforce configured so that only authenticated employees can open the labeling portal, run the job, and review results. Together they deliver managed labeling with human review loops and no exposure of the data outside the organization.Why the Other Options Are Wrong
Option A (Amazon Rekognition automatic labeling) offers no guarantee of accuracy against your own taxonomy and gives no employee-controlled review path, so it cannot satisfy "highest possible accuracy". Option B is not supervised learning at all — as one commenter put it, "This describes unsupervised learning, not supervised learning" — because letting the model infer its own labels defeats the purpose of a labeled training set. Option E uses Amazon Mechanical Turk, whose workers are outside the organization; as another voter noted, "its public crowdsourcing and data needs to be viewed only by employees", so it violates the access constraint regardless of cost or speed.Community Comment Notes
Every substantive voter converged on C and D, with bigdata2ml summarizing the split cleanly: "C - Supervised learning requires labels, Ground Truth can start labeling the images, D- Private team setup in Ground Truth can review the images". Another commenter dismissed the automated options because "Human validation is superior", and techrunnerisme added that "Supervised learning would require labelled data". Chelseajcole reduced it to the essentials: create the job with C, and with D "only employee to run". The lone vote for B appears to be a misread of the supervised requirement rather than a competing interpretation.Official Reference
Exam Strategy
When a stem says "Choose two", look for a service-plus-configuration pair: Ground Truth creates the job, and the workforce option decides who does the work. Any answer that routes confidential data to a public crowd or that skips labeling entirely contradicts an explicit constraint in the stem.
Frequently Asked Questions
Why is Amazon Mechanical Turk wrong for labeling employee-only images?
Mechanical Turk draws on a public, anonymous crowd, so confidential images would leave the organization's controlled workforce and breach the employee-only access requirement.
Could Amazon Rekognition label the images well enough instead of Ground Truth?
Rekognition automates detection using pretrained or custom models but gives no employee-reviewed ground truth for your specific taxonomy, so it does not meet the highest-accuracy labeling goal.
Related Analysis
Practice All MLA-C01 Questions
Access 115 questions with complete answers and detailed explanations.
View Full MLA-C01 Practice Test →