aloshai / imagerhub-backend Goto Github PK

Here I will collect and analyze free-to-use and available datasets.

I will be using yolov5 in Python to do AI integration. Since the datasets are too large, it is not possible to host this application on Heroku, glitch, or my server. I plan to use lighter packages like coco128 for this. Object365 looked fine to me, but 500GB+ is not something I can use.

My goal now is to integrate yolov5. (After making the Auth system) I have several plans, not sure which one to make.

My plans:

• Using Python on a contextual basis. So every time a new image is uploaded, I will create a sub_process targeting it and use python-shell.
• I will keep Python open like a microservice via Docker and pass the file paths of newly created Images to the console. In this way, I will access the necessary labels without the need for a constant build.
• When a new image is uploaded to the Python side using the MongoDB Document Watch function. I will use the image path and detect the labels and then update the document.

I haven't decided what to do now. In the future, I will prefer one of these steps.