I need to build an Image Similarity engine in Azure. I use Spotify Annoy library to build a Vector search DB. These vector db files are roughly near 10-15 MB.
The functionality is: User uploads an image. Embeddings for this image is generated using a pretrained model. These embeddings are used to find ANNs from the vector file. The ANN search gives the similar images name. Then the end user is shown these end images.
As of now I have created an Azure function which has the static Vector DB file inside the Azure function code files. And the web interface is hosted in a Storage container with the images (19000 total images).
My concern is that the static Vector DB file is big and everytime the function is created, it will have performance as well as pricing issues.
I am new to Cloud/Azure. Also can I do the same using Azure ML Studio?