Computer Vision
Convolutional Neural Networks with Deep Learning
Convolutional Neural Networks makes it possible for Deep Learning based models to understand images and videos and create various solutions involving images and videos.
Image Classification
Image Classification models classify images into one of distinct classes.
Some industry specific use cases for image classification are
Identifying product category from an image for e-commerce websites
Classifying sub-cellular protein patterns in human cells
Classifying houses into various categories for a real estate company
Object Detection
Object Detection models identify objects and their locations in an image.
Some examples of how this technique is used are
Identifying pedestrians, sign boards and other vehicles for a self driving car
Finding specific products from an image, counting their quantity and
identifying their position relative to other objects for retail planogram
automation
Identifying products from image to crop out extra part of the image for
e-commerce websites
Generative Adversarial Networks (GANs)
Generative Adversarial Networks (GAN) generate new images of object which they have been trained with. GANs can be combined with other techniques to create images of object with specific attributes.
GANs can be used in industries like
Fashion and apparel industry for creating new designs of jewellery, clothing,
watches, shoes etc.
Generating fictional faces for animation or other use
Face Recognition
Face Recognition models are able to detect a face from an image and match it with a dataset of existing faces to identify a matching face.
With cutting edge Deep Learning models, machine face recognition accuracy has reached human performance levels. Face recognition has many real world
use cases such as
Employee access control based on face authentication
Live face matching using video to authenticate users with a check for
liveliness
Identifying passengers by Kiosks/ Robots as Airports, Banks or other such
locations
Super Resolution
Super resolution is a technique used for Convolutional Neural Networks to be able to handle images with large resolution. Super resolution can be combined with any of the above models to allow them to handle images with higher resolutions.