In recent years, new deep learning technologies have made great advances in the field of computer vision, especially in image recognition and object detection.
That is why, we want to explain step by step its evolution. And for this, we rely on the viso.ai article.
Here we go!
During this decade, computer vision came to light, precisely when computer scientists tried to emulate human sight through computation.
Although the research went on for several more decades, the most advanced that was achieved at the time was the perception of common objects. However, it was very difficult to recognize multiple natural objects with variations in shape.
Known as the era of deep learning. Researchers achieved breakthroughs by training computers with at least 15 million images from the largest image classification dataset (ImageNet). All this was made possible by deep learning technology.
In all computer vision tests and challenges, deep learning demonstrated superiority over traditional algorithms.
Near real-time deep learning! Deep learning, a particular class of machine learning algorithms, is able to simplify the process of feature extraction and description using a multilayer convolutional neural network (CNN).
Thanks to ImageNet’s massive data, modern central processing units (CPUs) and graphics processing units (GPUs), deep neural networks bring unprecedented development of computer vision and achieve state-of-the-art performance.
Especially, the development of single-stage object detectors made deep learning AI vision much more efficient and faster.
During this year, the deployment of deep learning and Edge AI occurred. Today, CNN has become the de facto standard computing framework in computer vision.
In addition, numerous deeper and more complex networks have been developed for CNNs to provide near-human accuracy, in many computer vision applications.
Optimized, lightweight AI models enable computer vision to be realized on low-cost hardware and mobile devices.
Edge AI hardware, such as deep learning hardware accelerators, enable highly efficient edge inference for computer vision.
Finally, we can emphasize that Computer Vision technologies have evolved over time, going from the perception of common objects to almost human vision.
All this has been possible thanks to tests, studies and a lot of data capable of making the systems understand as they work.
In our next article we want to complement this information with 6 cases of uses of Computer Vision, do not miss it!