Defining the research or business problem
|
My passion is helping clients with data projects that have impact. In collaboration with different teams, I have successfully managed projects from conception to production.
The first and perhaps most important step in a project is defining the research or business goal. This means starting at the end and asking the question what the application should do and how. A thorough study of the use-case is therefore needed. Breaking down the question in different steps |
|
All projects can be broken down into different building blocks. Often, it includes one or more of the following steps:
- Data collection and annotation,
- Data exploration and visualization,
- Data pre-processing and quality inspection,
- Training and validation of machine / deep learning models,
- Experimentation with different parameter settings,
- Model deployment,
- Monitoring of model performance in production.
Use-case examples
Machine or deep learning can be used to classify characteristics of a photo without human intervention. For example, it can be used to identify the principal color of a photo. A color tag can amongst other caracteristics be used in an e-commerce site to facilitate product search.
Object detection is another area of applications. Please see below how it serves in spotting product defects. There are of course many more use case examples. Many use-cases aim to (i) verify product quality, (ii) to save user or production time or (iii) to assist in recognizing objects.
Please feel free to share your use-case!
