No Code Tools to Build Machine Learning Models

Dinakaran Sankaranarayanan


No Code Tools to Build Machine Learning Models

Machine Learning is complex. There is a steep learning curve and it is a pretty daunting task even for developers and others in the tech to get started without PhD in Maths, Statistics and Algorithms.

In one of my recent blog post , I have written about how we can use Machine Learning to implement various business problems without having to build any custom machine learning models from the scratch. AI services provided by cloud providers like AWS or GCP etc can help you get the job done for a few use-cases through the use of SDK and API.

While these work well in a limited context, there is a need to train machine learning models that are not generic and hence could not be well served by the AI services out-of-the-box. This is where the no-code ML tools come into the picture. Even though some of the no-code ML tools are targetted for building, training models for limited use-cases, it is still way easier to harness these for building Machine Learning Powered features. These tools mostly work on the Computer Vision and Natural Language Processing related sub-domains of Machine Learning.

Some of the no-code ML tools are :

1. Teachable Machine supported by Google

Teachable Machine is a browser-based no-code Machine Learning Tool that can help us to build and train machine learning models. Once trained, these can be exported and integrated with the functionalities into our apps.

This tool helps anyone to train ML model using images, sound and poses right from the comfort of a browser and webcam. The alluring aspect of this is the easy-to-use nice interface that can help to easily build models with minimal effort without the need for any coding.

2. supported by Microsoft

Lobe is another ML tool that helps us to train models similar to Teachable Machines. Instead of a web browser, there is a desktop application that is available and offers a lot of features for data gathering, training, error and testing. At present, only images are supported, but other types will be supported in the near future. The site is really nice and the short tour and examples help to get started pretty quickly.

What next?

One of the reason companies are really trying to get Machine Learning accessible to all is because of the innovation that can be fuelled on top of these use-cases. Some of the hard problems that are associated with Machine Learning are getting simplified through a set of simple tools that can help anyone get started. Also, the tough architectural challenges are also being solved by building custom products and the best practices /processes around running ML models in production are also being widely discussed and evolved. It should be very interesting to see how the Machine Learning space evolves in the next few years and how these new tools will be leveraged by builders throughout the world that can help to solve some pretty interesting problems.