Three applications deep learning and computer vision can improve

Utilizing deep learning and computer vision can improve image classification, object segmentation and image reconstruction.

By AIA March 11, 2020

You’ve probably heard about deep learning and computer vision and you’re likely aware that they have something to do with artificial intelligence. If you’re like most people, you’re not quite sure what that means. Are there practical uses for deep learning and computer vision? Short answer: Yes.

Artificial intelligence isn’t just for sci-fi movies anymore. Deep learning and computer vision are powering complex systems and are making smart decisions right now. Let’s explore three computer vision applications with deep learning intelligence.

1. Image classification

One of the most common uses for deep learning and computer vision is classifying images. Complex analyses are performed to label an entire image or photograph. Some examples include labeling disease on a medical scan, transcribing handwritten notes, and identifying faces in a photo.

Deep learning algorithms compare an image to images in a dataset and search for similarities. The datasets for image classification include previously labeled images of all kinds of variations such as photos of various dog breeds labeled as “dog” and photos of sedans, coupes, and convertibles labeled as “car.”

2. Object segmentation

Also known as semantic segmentation, object segmentation draws a line around each object detected in an image. Object segmentation then identifies the specific pixels in the image that belong to the object.

Again, datasets are used for object segmentation. Separate objects that are commonly detected in computer vision applications with deep learning include people, vehicles, and structures. Object segmentation is useful to an autonomous vehicle to avoid pedestrians, vehicles, and road hazards.

3. Image reconstruction

Deep Learning and computer vision can even help to restore parts of an image that are missing or that have been corrupted. Images that may be missing data may include photographs that have been damaged. It may also include surveillance images when a person was obscured by another person or an object.

Image reconstruction can reconstruct old, damaged photographs and movies with the help of a dataset that includes an image and also a copy of that image with data missing. Deep learning algorithms fill in the gaps to create a whole image. Perhaps some surveillance footage only captured half of an assailant’s face. Deep learning can generate a complete image.

This article originally appeared in Vision Online. AIA is a part of the Association for Advancing Automation (A3), a CFE Media content partner.

Related Resources