Do research on the database

behzadshomali / Image-Describe-Pipe

This app outputs the name, coordinations, sentiment of each extracted face, and besides a brief description of the scene's context for each input image.

2 stars 0 forks source link

Do research on the database #3

Closed behroozomidvar closed 3 years ago

behroozomidvar commented 3 years ago

[x] What database architecture is required for the problem?
[x] Which DBMS to use? Why?
[x] What are the attributes and their data types?

behzadshomali commented 3 years ago

@behroozomidvar according to "DeepFace" documentation, to be able to apply face recognition we have to pass the path of the directory where we stored our images. By having in this mind, I wonder if we really need a DBMS. What is your opinion (or solution) in this regard?

behroozomidvar commented 3 years ago

DBs are not only for storing images but also for user sessions and login info. Also, image meta-data is not limited to its address but we might want to enrich it with other stuff as well. Anyhow DBs are not necessary for short-scale projects, and they're inevitable for medium to large ones.

behzadshomali commented 3 years ago

@behroozomidvar according to our needs, a single tabular DBMS can fulfill us. In other words, I think we won't need NoSQL DBMSs.

behzadshomali commented 3 years ago

@behroozomidvar after a brief search, I found out that PostgreSQL can be one of our probable choices, cause:

it's free
it supports JSON which can simplify our further works in many terms
it supports many pre-defined functions and also another choice can be MariaDB which fairly fast and stable.

behroozomidvar commented 3 years ago

I've worked with Postgres before, and I'll definitely approve. Not sure about MariaDB. Also, have you checked MySQL and SQLite options?

behzadshomali commented 3 years ago

@behroozomidvar that's great! I've also worked with SQLite, but I really adore trying new stuff, but it's a good choice. Frankly, I didn't check MySQL because among all the DBMSs I checked I found Postgres much interesting and handy so I didn't keep on searching.

behroozomidvar commented 3 years ago

It makes perfect sense @behzadshomali. Hands down when it comes to Postgres. The two others are often synonyms to fast or dirty coding. Hence let's go with the former.

behzadshomali commented 3 years ago

In terms of DB attributes I found following properties but I'm sure that these are not all the attributes we should care about:

User images: each person have a unique collection of images of his relatives, friends, or the people he plans to look for in his future uploaded images. To respect for user privacy, I suggest that instead of storing user images, store the features extracted from the images.
User info: information such as username, password, email, name, age, and etc. .
User login logs

@behroozomidvar what's your opinion? what other attributes should be considered?

behroozomidvar commented 3 years ago

Let's play incremental. For now, all is good. 👍