Introduction to Image Processing

BigML’s upcoming release on Wednesday, December 15, 2021, will be presenting a new set of Image Processing resources to the BigML platform. In this post, we make a quick introduction to the basic concepts before we move on to the remainder of our blog post series that will give you a detailed perspective on composite sources, image labeling and classification, together with workflow automation, and use case examples. Finally, we’ll cap the series with a behind scenes peek into how BigML’s Image Processing implementation works.

Image Processing Defined

Image Processing is an interdisciplinary technical concept with flavors such as computer vision or machine vision depending on the context within which image data is being put to use to gain actionable insights. While our impending release sets the foundation for many types of Image Processing capabilities to follow, when we refer to Image Processing we’re currently concerned with determining whether an instance of image data contains specific features. 

The underlying algorithm enabling this type of image analysis is known as Convolutional Neural Networks (CNNs), and combined with advances in GPUs it has been responsible for the dramatic improvements we have witnessed in the ImageNet challenge in the past decade. Today, Machine Learning-driven image understanding in many domains has reached performance levels close to or even exceeding that of humans, especially for highly specialized problems. One of the reasons why CNNs have become popular is that they automate the learning process from image datasets as opposed to traditional algorithms that require hand-engineered features. However, because CNNs also tend to be quite compute heavy compared with traditional ML techniques, so there’s a trade off. The practitioner has to decide whether the additional training time and complexity is worth the investment on a per use case basis. BigML’s comprehensive set of ML algorithms give you that choice and the chance to compare and contrast.

Image Processing Use Cases

With the advent of commercial computer vision systems and smartphone technology, an abundance of image data is now all around our workplaces and in our daily lives as consumers. As a result, many businesses are sitting on top of valuable image data that goes unexploited despite the immense potential they represent for many industries. BigML Image Processing is built from scratch to catalyze the mainstream adoption of image-based smart applications by making it is easier than ever to solve a wide variety of visual and/or mixed data type use cases while offering the built-in traceability and auto-scaling that it has come to be known for. Just to name a few:

• Manufacturing: Today’s manufaturing operations are highly automated with industrial IoT and robotics as driving forces. Analyzing image data opens up new possibilities in detecting product defects early in the process thus reducing waste as well as enabling predictive maintenance to expand the productive life of key assets and equipment.
• Transportation: Infrastucture services players such as toll road operators resort to image data when it comes to automating decisions ranging from dynamic pricing of managed lanes based on traffic patterns to monitoring assets and tracking vehicles as identifies by their license plates.
• Healthcare: Medical professionals such as radiologists have come to rely on image data in the form of X-rays, CT scans and MRI for decades. Now Machine Learning systems can help with better and more consistent disease diagnosis and patient recovery tracking.
• Defense: Drones and autonomous defense systems heavily rely on the ability to interpret fast moving visuals and video footage to assess, contain and retaliate against a variety of manned or unmanned threat vectors. In the theatre of combat, success is often measured by concise decision making in seconds.
• Retail: Image data plays a key role in visual search for hard to describe products such as apparel, surveillance and security in stores guarding against theft, managing and tracking inventory and many more use cases that would otherwise be impractical.

What makes BigML Image Processing different?

While there are a variety of open source and commercial image processing libraries and frameworks available to Machine Learning practitioners, BigML’s unique implementation offers the following advantages that can boost productivity and accelerate time to market in building and deploying smart applications.

All the Tasks for Generating Insights from Image Data on a Single Platform

From labeling to inference, evaluation, and predictions all the relevant tasks for generating insights from your image data and automating business processes can be performed with BigML Image Processing. Since BigML treats images as any other data type on the platform, BigML users can easily use image data alongside text, categorical, numeric, date-time, and items data types as input to any type of Machine Learning model, both supervised and unsupervised. With this release, users with different levels of proficiency avoid having to learn yet another tool to make sense of their image data.

Streamlined Image Dataset Management with Composite Sources

A composite source is a collection of component sources supporting multiple data types including individual images. Composite sources save time and prevent errors because users can incrementally add more data to them and use the built-in image labeling while preserving immutability along the way.

Comprehensive Feature Extraction Options to Feed Any Algorithm

Through multiple feature extraction configuration options, BigML gives you fine control over what the algorithm “sees” in your image data. As part of pre-processing, a collection of images are automatically transformed into rows of tabular data that any model type can handle with ease. If the user prefers Convolutional Neural Nets (CNNs), BigML Image Processing can still help make the resulting predictions more interpretable despite the fact that CNN’s have a reputation for being less interpretable.

Pre-Trained CNNs for Classification and Regression

BigML also allows the user to pick from five different pre-trained Convolutional Neural Nets to build better models faster by leveraging Transfer Learning. These pre-trained models are based on industry-standard datasets containing millions of images and thousands of classes.

Want to know more about Image Processing?

If you’d like to learn more about how Image Processing works, please bookmark the release page as it will be continuously updated to include links to the upcoming blog posts, the latest BigML Dashboard and API documentation, the webinar slideshow, as well as the full webinar recording past the official launch date. And remember to register today for the FREE live webinar on Wednesday, December 15 at 8:30 AM PST / 10:30 AM CST / 5:30 PM CET!