Object detection, segmentation and classification are the building blocks to address several complex computer vision challenges. Object detection helps to identify an object in the image, forms a rectangular boundary and creates a bounding box to narrow down the object. Image segmentation then identifies the object with all the curves, lines and exact shape of the object. This helps in more granular and finer identification as against simple object detection. Object classification helps classify a particular object into a class or subclass. For example, classifying a vehicle into a car or an airplane and further sub classifying the brand into Audi, BMW, etc.
Object detection, segmentation and classification help in the healthcare sector by identifying and narrowing tumor regions and further classifying the tumor as malignant or not.
These also help to establish various insights from images by classifying them to categories, segmenting them for specific information, or extracting any image attributes.
Infosys partnered with a large global energy company to identify faulty cables based on the picture sent from the site of failure, which allowed them to send the right engineer to the site to fix the cable. This helped the company in saving costs from sending engineer to fix wrongly reported cable problem.
Infosys worked with a large global retailer to extract and classify information from digitally scanned product art (SmartArt) so that the information was extracted correctly and could be further classified as contents, ingredients, instructions, etc. to make the information available on multiple channels for regulatory and compliance purposes.
There are several interesting possiblities emerging from applying AI to videos, such as generating video captions, video highlights, content moderation, span of brand coverage, surveillance, and people or object tracking.
As part of a prestigious global tennis tournament, using various CV-based algorithms, Infosys extracted various game insights to create highlights from various events, such as seeing players waving to the crowd, extracting the score from the video feed, recognizing players and determining the length of the time a particular advertisement or brand was featured in a video.
Similarly, for a large railroad company in the U.S., various assets spread across geography were identified and counted from a streaming video feed obtained using a train-mounted camera.
Some of the new problems we are working on with clients in the CV space include handwriting recognition in the context of know-your-customer forms that are written manually and need to be digitized, activity and pose recognition in a video, video synthesis, video summarization, and image captioning by leveraging state-of-the-art AI models and techniques such as 3D object detection, generative networks and single-shot learning.
To keep yourself updated on the latest technology and industry trends subscribe to the Infosys Knowledge Institute’s publicationsCount me in!