The Google Vision API web detection feature can be used only to return the contents of the image as they relate to the Google Knowledge Graph, as well as the image's relation to other pages and images on the web.
Based on this, it is not currently possible to use it by specifying your own database of images; However, I suggest you to use the Send Feedback button located at the lower left and upper right corners of the Detecting other features public documentation in order to notify to Google about this desired functionality.
You can upload your images to google cloud and do the job. You have 2 choices: Google product search or AutoML. With product search you can compare an image with a set of images of products. You will find product form image. You can also use AutoMl, upload some images and train them. You can predict images with AutoMl API in this case.
