Practice Test 2 | Google Cloud Certified Professional Data Engineer | Dumps | Mock Test
You want to build a system which uses a machine learning, image recognition model to detect customers’ faces entering a retail shop, and based on the knowledge base it will return whether the customer is a new, returning or loyal customer. You are building the model using AutoML Vision. After training the model and testing it, you find the model’s accuracy is lower due to overfitting. How can you solve this?
A. Images used should be taken from the same exact angle and resolution.
B. Instead of manually splitting samples to training and testing sets, allow AutoML Vision to split the sample set.
C. Samples used for training should be covering true positives only.
D. Images used should be taken from different angles, resolutions and points of view.
Answer: D.
Description:
Google Cloud provies a machine learning service called AutoML to quickly build models for you. AutoML Vision is one of its products which you can start with a training set as little as a dozen photo samples and AutoML takes care of the rest.
While iterating on your model, if the model’s quality levels are not up to expectations, you can go back to earlier steps to improve the quality:
- AutoML Vision allows you to sort the images by how “confused” the model is, by the true label and its predicted label. Look through these images and make sure they’re labeled correctly.
- Consider adding more images to any labels with low quality.
- You may need to add different types of images (e.g. wider angle, higher or lower resolution, different points of view).
- Consider removing labels altogether if you don’t have enough training images.
- Remember that machines can’t read your label name; it’s just a random string of letters to them. If you have one label that says “door” and another that says “door_with_knob” the machine has no way of figuring out the nuance other than the images you provide it.
- Augment your data with more examples of true positives and negatives. Especially important examples are the ones that are close to the decision boundary (i.e. likely to produce confusion, but still correctly labeled).
- Specify your own TRAIN, TEST, VALIDATION split. The tool randomly assigns images, but near-duplicates may end up in TRAIN and VALIDATION which could lead to overfitting and then poor performance on the TEST set.
Once you’ve made changes, train and evaluate a new model until you reach a high enough quality level.
Source(s):
Cloud AutoML Vision – Evaluating Models:
Comments are closed, but trackbacks and pingbacks are open.