Product detection API from Custom Vision & Computer Vision

Question

I want to use product detection api from custom vision and computer vision, here are few questions i have on them

1- Say we train the model to identify X magazine based on a set of images, now after we want to do prediction the version of magazine X will be a different one as a new version is released every week, will the model still be able to identify the new version of magazine X?

2- I have a dataset of 500 images(300 for training, 100 for validation,100 for prediction) what is the training time I can expect from the model?

Thank you!

Answer 1

Hi @Rishi Khemka

Thank you for using the Microsoft Q&A platform.

Here are the answers to your questions:

1. If you train the model to identify X magazine based on a set of images, the model should be able to identify the new version of magazine X, as long as the visual characteristics of the new version are similar to the ones in the training set. However, if the new version of magazine X has significant visual differences from the previous version, you may need to retrain the model with new images that include the new version of magazine X or try implementing incremental training (optional), whichever fits your scenario. Note that the effectiveness of incremental training depends on the similarity between the new images and the existing training set. If the new images are significantly different from the existing training set, you may need to manually - Test and retrain a Custom Vision model with the new images.
2. The training time for the model depends on various factors such as the complexity of the model, the number of images in the training set, and the compute resources & number of transactions made for training. It's difficult to estimate the training time without knowing more details about your specific scenario. However, you can use the Azure Custom Vision pricing calculator to get a rough estimate of the cost and time required for training your model based on your specific requirements.

I hope the provided information helps! Thank you.

---

Please don't forget to click Accept Answer and Yes for was this answer helpful.