Exam AWS Certified Machine Learning Engineer - Associate MLA-C01 topic 1 question 58 discussion

It's B. Did you guys clearly understand the question? "The model performed well during training and validation. However, the model is underperforming in production because of variations in the quality of the images from various cameras." https://aws.amazon.com/blogs/machine-learning/prepare-image-data-with-amazon-sagemaker-data-wrangler/ Corrupting an image or creating any kind of noise helps make a model more robust. The model can predict with more accuracy even if it receives a corrupted image because it was trained with corrupt and non-corrupt images.

It's C - The problem is described as "variations in the quality of the images from various cameras" - this primarily suggests issues with lighting, contrast, and image quality variations. Not B, because the corrupt image transform with impulse noise option adds random noise to images, which wouldn't help with camera quality variation and could potentially make performance worse.

Recreate the training dataset by using the Data Wrangler corrupt image transform. Specify the impulse noise option. ✅ Adds noise to images, simulating low-quality or distorted camera feeds. ✅ Fast — can be done with existing data. ✅ Makes model more robust to variations in image quality. ✅ Best for improving accuracy in the least amount of time.

Corrupting an image or creating any kind of noise helps make a model more robust. The model can predict with more accuracy even if it receives a corrupted image because it was trained with corrupt and non-corrupt images. https://aws.amazon.com/blogs/machine-learning/prepare-image-data-with-amazon-sagemaker-data-wrangler/

Option B focus on impulse noise targets noisy images, not quality variations like contrast or lighting. Option C focus on contrast enhancement fixing lighting and clarity issues, directly addressing camera quality variations.

https://aws.amazon.com/blogs/machine-learning/prepare-image-data-with-amazon-sagemaker-data-wrangler/

Enhancing image contrast can help standardize the quality of images from various cameras, making the model more robust to variations in image quality