Computer Vision and Auto ML – Introduction
Computer vision is rapidly gaining popularity through AI automation and remote monitoring. With the new demands and challenges presented in the past few years, we’re seeing a massive impact on companies from all industries when it comes to computer vision. So what is computer vision, and how can you start using image data in your models?
What is Computer Vision?
Computer vision is an interdisciplinary field of science that refers to the ability of a computer to see and understand images. Much like the human vision system, computer vision enables computers to derive meaningful information from digital images or videos and then take actions or make predictions based on that information. Thanks to advances in AI, the field has been able to take great leaps in recent years and has even been able to surpass humans in some tasks related to detecting and labeling objects.
Where is Computer Vision Used?
Businesses are increasingly adopting computer vision in a number of essential industries. In the healthcare sector, computer vision is used to improve the cancer screening process and detect tumors with higher speed and accuracy. In agriculture, computer vision is being used for everything from quality, size, and shape classification to crop yield prediction and disease detection. Cruise and Tesla are among many companies leading the venture into autonomous cars; self-driving systems that use computer vision paired with AI to detect movements in 360 degrees to watch out for other vehicles, pedestrians, and cyclists.
Barriers to Entry
So if computer vision can be of value in such a variety of use cases, why is it not more ubiquitously implemented?
Well, until recently computer vision only worked in a limited capacity. Image data differs from other types of data because of the large size of the files, and consequently the size of the datasets. Labeling and preparing large datasets make training a new model a resource intensive, costly, and often time-consuming task depending on the limitations of your machine and availability of GPUs. On top of all this, selecting, training, tuning, and monitoring a model requires some technical coding and machine learning expertise. As a result, these challenges present significant barriers to entry for engineers at any experience level.
Automated Machine Learning
Fortunately, there has been a proliferation of applications and platforms that cater to removing obstacles to implementing computer vision. Datature, for example, offers a managed infrastructure to build computer vision applications entirely without code. They even offer an annotator feature to label images quickly and collaboratively with IntelliBrush. Alternatively, DataRobot’s Visual AI platform offers a centralized environment for your computer vision and multimodal data projects; all the way from data prep to enterprise grade operationalization and monitoring. These types of solutions allow users to streamline their ML stack and validate models, without needing a single lesson in OpenCV. Thanks to automated machine learning, you don’t need to be a computer vision expert to start leveraging image data in your models right away
Figure 1 – Through DataRobot’s UI, just drag and drop a zip file of images into a new project, choose a target, and start training.
Conclusion
Computer vision is a field that seeks to understand digital images to automate or outperform on visual tasks. Typically, building a computer vision system requires a combination of theoretical knowledge and technical skill. However, AI powered Computer Vision and the potential value it can return, is and should be available to everyone. In Part 2, we will further examine DataRobot, and demonstrate how to build a simple computer vision model using AutoML with Visual AI.
