AIF-C01 Question 45
Select 3A company is building an AI-powered application for image recognition but has limited resources and expertise in training machine learning models. Which of the following approaches should the company consider for quickly implementing a solution?
- A
Use an open-source pre-trained model available on a platform like TensorFlow Hub or PyTorch Hub.
- B
Train a custom machine learning model from scratch using a large labeled dataset.
- C
Leverage an AWS AI service like Amazon Rekognition for pre-built image recognition capabilities.
- D
Use Amazon SageMaker to automatically fine-tune a pre-trained model with the company's specific dataset.
Show answer and explanation
Correct answers: A, C, D
Explanation
For a company with limited resources and expertise, leveraging pre-trained models or pre-built AI services is often the most efficient and cost-effective approach. Options like open-source pre-trained models, AWS AI services (e.g., Amazon Rekognition), or fine-tuning pre-trained models using services like Amazon SageMaker provide quick implementation paths without requiring extensive expertise or resources. Training a model from scratch is generally not recommended in this scenario due to the complexity and resource requirements.
- A. Correct.
Using an open-source pre-trained model is a valid approach, as it allows the company to quickly leverage a model that has already been trained on large datasets. This saves time and computational resources.
- B. Incorrect.
Training a custom machine learning model from scratch would require significant expertise, resources, and a large labeled dataset. This is not an optimal approach for a company with limited resources and expertise.
- C. Correct.
Amazon Rekognition provides pre-built image recognition capabilities, which is an excellent choice when you want to avoid managing any infrastructure or custom training.
- D. Correct.
Using Amazon SageMaker to fine-tune a pre-trained model is a practical and efficient way to adapt an existing model to the company's specific needs, without training a model from scratch.