Exam AI-102 topic 2 question 37 discussion

First step should be Custom Vision. Then Upload sample images / Label / Retrain / Publish Custom classifiers have to go through custom vision, obviously

Given that the prerequisite is to minimize development effort, the first step should be to use the Custom Vision Portal, not Visual Studio, a that requires more development effort than a GUI, such as the Vision Portal.

I think from now on the first step shold be Vision Studio, which if I get correctly is the new Custom Vision Studio https://learn.microsoft.com/en-us/azure/ai-services/custom-vision-service/concepts/compare-alternatives

From the Custom Vision portal, open the project. Upload sample images of the new products. Label the sample images. Retrain the model. Publish the model.

From the Custom Vision Portal, Open the Project. Upload Sample images of the new products. Label the Sample images. Retrain the model. Publish the model.

1. From the Custom Vision portal, open the project 2. Label 3. Upload 4 .Retrain 5. Publish

Will this question be on the actual exam?

So I will choice, 1. From the Custom Vision portal, open the project 2. Upload sample images of the new products 3. Label the samples images 4. Retrain the model 5. Publish the model labelling after uploading sample images.

To add new products to the classifier while minimizing time and development effort, you should perform the following actions in sequence: From the Custom Vision portal, open the project. Upload sample images of the new products. Label the sample images. Retrain the model. Publish the model. This sequence ensures that the new product images are properly added, labeled, and incorporated into the existing model, and that the updated model is made available for use by your application.

Final Answer: 1. From the Custom Vision portal, open the project 2. Label the samples images 3. Upload sample images of the new products 4 . Retrain the model 5. Publish the model

4) Retrain the Model: Trigger the retraining process for your custom classifier. This step involves: Using the labeled samples to train the model. Fine-tuning the existing classifier with the new data. 5) Publish the Model: Once the retraining is complete and the model performs well on validation data, publish the updated classifier. The published model will be ready for inference in your application, allowing it to identify the new products.

Your proposed sequence of actions for adding new products to the classifier is a good start! Let’s refine it a bit to ensure it aligns with best practices: 1) Open the Custom Vision Project: Begin by accessing your project in the Custom Vision portal. This is where you’ll manage your custom-trained classifier. 2) Label the Sample Images: Next, label the sample images you’ve collected for the new products. Assign appropriate tags or classes to each image based on the product category. Proper labeling is crucial for effective training. 3) Upload Sample Images: Upload the labeled sample images to your project. These images will serve as the training data for your classifier. Make sure you have a diverse set of samples to represent different variations of the new products.

According to chat GPT, the correct sequence is the following : 1. From the Custom Vision portal, open the project 2. Label the samples images 3. Upload sample images of the new products 4 . Retrain the model 5. Publish the model Labeling the sample images before uploading them is crucial because it helps in structuring and organizing the dataset appropriately.