How to apply MLOps to Computer Vision? Introducing CVOps

기사 원문

Abstract

You may be familiar with Devops that developing and testing these applications using a Continuous Integration and Continuous Delivery pipelines.

Since MLcycle is similar to software development cycle, MLops was developed.

MLops has several steps. 1) Framing the business objectives, 2) Searching for the relevant data, 3) Preparing and processing the data (Data Engineering), 4) Developing and training the Machine Learning model, 5) Building and automating a machine learning pipeline, 6) Deploy the model via static or dynamic deployment.

In the field of CV, prototyping Computer vision models is easy, but building an integrated ML system is difficult.

To achieve MLops, we commonly use some ML platforms, that might be end-to-end.

Thoughts

Although I am beginner and just busy identifying tasks, realized that prototyping Computer vision model is tiny part and there are many other things. I felt the need to study about MLops. I will search ML platforms and apply it to our tasks.