Vertex AI at scale

[SmSpt]

Hi, 

I have been playing around with a combination of a Vertex AI object detection model and the Vision API to read text from images. So far the maximum amount of images I have run my model on is around 3000 images costing around €0,0019 per image and taking hours to finish. 

I am looking at a project which would require me to run around 1.5 million images through the process. What concerns me is how long that would take time-wise and the cost of a full run. Therefore, Im looking at  the sku options to see whether I should upgrade, and whether this would increase my performance significantly. Keeping an eye on price is there somewhere who has worked with projects of such scale, and how they set it up. 

Thanks!

[nceniza]

Google has a pricing calculator that will let you estimate the cost depends of resources, feature to be used, etc etc : https://cloud.google.com/products/calculator/

Additional infomration is provided in the "?" icons to help you decide to identify these factors:

For estimated time, I do not believe there is a calculator for that, you will have to rely on your initial run ( given that you are using the same resources) and calculate the estimate time for 1.5 images, Alternatively you can contact sales to help you decide in your use case and estimates: https://cloud.google.com/contact