From the looming reality of fully autonomous vehicles to a farmer in Japan using TensorFlow to sort cucumbers, computer vision machine learning is becoming an increasingly accessible and ubiquitous phenomenon. As a service provider in this space, Sama has a unique view into the breadth of computer vision applications. While certain applications of machine learning are prevalent in mainstream media (e.g. autonomous vehicles), there are equally interesting and lesser-known applications that are similarly revolutionary.
Satellites orbiting the earth have been collecting comprehensive image data sets of the globe for decades. With recent advancement in computer vision, these images can be efficiently and automatically combed for valuable information. By applying deep learning models to sets of satellite imagery, companies like Orbital Insight can extract data on global surface water levels, informing communities, planners, and policymakers in making critical decisions about water resources.
Figure 1 San Luis Reservoir. Left: Raw image. Middle: Landsat 8 water mask. Right: Orbital insight water mask. Orbital Insights, 2015
Image analysis in the medical field is a time-consuming process; using computer vision machine learning to automate portions of this analysis could help make the process more efficient and less costly for the general public. A bottleneck to progress in this field has been the lack of training data, which requires specialists to create (though there’s evidence that specialists can be substituted by pigeons!) and the data is typically confidential. The availability of large datasets with diagnostic labels, such as the Alzheimer’s Disease Neuroimaging Initiative (ADNI), has paved the way for new applications, using computer vision machine learning to provide early diagnosis of Alzheimer’s disease.
The advancement of drone and computer vision technology has both lowered the cost of gathering huge sets of aerial imagery, and also increased our ability to intelligently extract information from that imagery. These technologies together can help farmers identify crop diseases or predict crop yields, automating a process where the alternative was manual inspection. In collaboration with Paul Allen towards his Great Elephant Census project, Sama facilitated the creation of a computer vision machine learning model used to track and count elephants as a part of anti-poaching efforts.
Ultimately applications of computer vision machine learning are limited only by human imagination. We look forward to seeing what applications this evolving technology will create, and are thrilled to count ourselves as part of this thriving community.
If you want to learn more about the future of practical computer vision, please join us at the Embedded Vision Summit, being held in Santa Clara from May 1-3. It’s the preeminent event for anyone adding computer vision capabilities to their products. We’ll be participating in the Vision Technology Showcase and invite you to stop by our booth.
The event is designed to inspire you to use vision technology in new ways and to empower you with the practical know-how you need to integrate vision capabilities into your products. Over three days and four tracks, you’ll meet innovators, luminaries, and colleagues in this fast-growing field.