Answer-first summary for fast verification
Answer: Use Cloud Vision AutoML with the existing dataset.
Given the large number of unique components (750) and the substantial dataset (1000 examples per component), along with the tight deadline for creating a proof-of-concept, utilizing Cloud Vision AutoML with the existing dataset is the most efficient and practical solution. Cloud Vision AutoML, a Google Cloud machine learning tool, allows for the quick training of custom image recognition models. Reducing the dataset size (Option C) could impair model accuracy, while relying on custom labels (Option B) may not be universally applicable. Training a model from scratch using transfer learning (Option A) would not be feasible within the given timeframe.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
You work for a manufacturing company that sources 750 different components from 750 suppliers. Your team has a labeled dataset with approximately 1000 examples of each unique component. The goal is to develop an app that helps warehouse workers identify incoming components by analyzing photos. You need to create a Proof-Of-Concept for this app within a few working days. What is the best approach to achieve this?
A
Train your own image recognition model using transfer learning techniques.
B
Use Cloud Vision API and provide custom labels as hints for recognition.
C
Use Cloud Vision AutoML, but reduce the size of the dataset by half.
D
Use Cloud Vision AutoML with the existing dataset.
No comments yet.