AI Glossary Header

Go-To AI Glossary and Computer Vision Definitions

haana -

It’s important to understand the definitions of each industry phrase but also the distinction between similar ideas. We created a quick go to guide to clearly explain each term. Learn with the below glossary to help you understand and boost your AI vocabulary.


    • AI: Artificial intelligence is a robotic or computer system that allows for an automation or tasks that would normally require human action or intelligence.
    • AI Accelerator: Specialized type of computer hardware system that can complete data-intensive and sensor driven tasks. Designed to help quicken delivery or processing of AI applications, such as artificial neural networks.
    • AI Model: A trained intelligence system using data, hyperparameters, and human intelligence to replicate a result the algorithm or expert has concluded. This model can be reused and shared.
    • AI Platform: A framework that helps developers create their own AI models. This software can be used for not only building but deploying and managing the AI application.
    • AI Training: A method of training AI models to ensure high level of efficiency in the task at hand. Once adjusted parameters are added, the AI can continue to improve its learning.
    • Algorithm: Outlined mathematical step-by-step set of instructions and  components implemented on the computer used to solve a problem or perform a task.
    • Artificial Neural Network (ANN): Inspired by the human brain’s neural network, this process uses images as related to the human visual system. Whatever images are inputted into the system is what it will be trained to understand and try to find.
    • Annotation: Comments or notes labeling the input data usually used in code to help explain the section. For the AI, the data frames need to be noted correctly including labels such as location, shape, or time.
    • API: The  application programming interface is a set of operational software components and computer functions that allow for data access and communications from an application creation  to accessing another. An AI API can refer to access to artificial intelligence tools and functionality. 
    • Backpropagation: An algorithm to help retrospectively detect error functions using gradient descent. Usually used with feedforward ANN.
    • Big Data: Delivers expansive, but raw datasets to help fuel information to complex systems, without the limits of an AI. Big data may have issues with quality of data or timing and can be analyzed with AI. 
    • Cloud Computing: This  occurs when data is gathered and transmitted to a data center for processing. The transmitted data is also stored in the central server and all devices that need to access this data or use applications associated with it must first connect to the designated server. Usually internet connection is needed for the devices to compute to the cloud.
    • Computer Vision:  A subfield AI to allow a computer to run high-level intelligence tasks and detection in processing and analyzing frames of videos, real-time.
    • Convolutional neural network (CNN): A type of ANN used with image recognition and processing data, down to the pixel level.  It is highly accurate when classifying because it creates an output based on a hierarchy of networks or funnel.
    • Demo: A demonstration of the features and capabilities of an AI platform, or of artificial intelligence in general.
    • DeepFake: An altered video usually of a person in place of the original image to make it appear authentic.
    • Deep Neural Networks: A type of ANN that features multiple layered model between the input and output layers. It has a non-linear relationship which allows for feature extraction from even raw input. Lower levels may include line detection or outlines whereas higher level layers are concept oriented.
    • Edge Device: This is a physical piece of hardware that is used to connect and flow data usually between networks. These devices may have limits in updating and becoming outdated.
    • Edge Computing: Is the process of capturing and processing data close to the source, either in a device or via an on-premise server.
    • Edge AI: The use of AI and machine learning algorithms running on edge devices to process data on local hardware, rather than uploading it to the cloud. Perhaps the greatest benefit of Edge AI is faster speeds (since data does not have to be sent to and from the cloud back and forth), enabling real-time decision-making.
    • Edge AI Proxy: This is a protocol that allows users to ingest multiple video streams. Our edge proxy has a common interface for conducting AI operations on or near the edge and it’s easy-to-use collection mechanism from multiple cameras onto a single more powerful computer
    • Facial Recognition: This is a type of biometric used to identify and confirm various facial landmarks of a specific person or people. In AI, facial recognition is used for real time video surveillance or facial authentication. 
    • GPU: The graphics programmable processing unit and is the fastest way to provide graphic and real-time images. Though a GPU can process a lot of data simultaneously and be useful for machine learning and AI, it can be costly and resource-intensive.
    • Hash: A common and fast way to vectorize features. The function is used to index the original value and then can be referred to or compared to a later data point.
    • Image Quality Control: A key quality control of images or videos by using AI to perform series of check and tasks to ensure the integrity and truth of the data. For example, some tools may be able to help with real-time video blurriness, or identifying deepfakes.
    • Image Recognition: Quite like facial recognition, this is a subfield of computer vision that is trained to high-level recognize the contents of an image or frame.
    • Inference: This is involved with a specific AI model to predict unknown, incoming data while using different training sets.
    • Infrastructure as a Service (IAAS): An industry that focuses on providing infrastructure tools mainly with basic computing such as server logic, cloud computing, network resources, etc.
    • Labeling: Vital part of AI to identify different object and features in the dataset. It is applied to the raw data to identify the file, but also later on to provide context.
    • Machine Learning: A concept that algorithms can be trained to learn as humans might. With the input of data and data training ML provides automatic improvements and learning without human intervention.
    • Metadata: Data that describes other data usually in more explanatory form. This may be technical, administrative or descriptive information.
    • Object Recognition: Quite like image recognition, this seeks to recognize one or more objects contained in an image or video frame. It can detect beyond the context identifying each given object displayed.
    • Pattern Recognition: Classification ability to automatically identify specific data and characteristic engagements.
    • Segmentation: A way of partitioning images and boundaries into multiple parts when the data can be expansive. For example, medical uses in identifying multiple cell structures by outline.
    • Software Development Kit (SDK): A set of tools including data libraries, documentation, and code needed to help build a specific software or platform.
    • Structured Data: Highly organized system of data into a tabular structure that is easily formatted. This allows for higher level quick analysis.
    • Supervised Learning: A limited way of learning based on the specific labeled input data. This allows for high-level accuracy because it is trained for a particular output.
    • Tagging: Similar to annotations, but this labelling process only uses one word for the input data during the AI training. For example a traffic AI may use tags such as: truck, sedan, bicycle.  
    • Unsupervised Learning: A process in which data patterns are identified without being given a specific classification of data points. It will construct its own model of the data and discover hidden patterns. This process is great for testing AI. 
    • Video Analytics: Using computer vision and machine learning to real-time and automatically detect contents of the video by parsing it into image frames. Depending on the AI, it can object or facial recognize along with allowing the ability to monitor or alert. 

More great articles

Chrysalis Cloud Face recognition with Python SDK

Face Recognition Using Chrysalis Cloud Python SDK 101

We’re going to show how to do face recognition in the cloud using Chrysalis Cloud. We break it down for…

Read Story
Chrysalis Cloud is the easy solution

New to Chrysalis Cloud? Check out the demo video.

If you're taking advantage of our one month free trial, then take advantage of this 15-min demo video. Click below…

Read Story
AI to aid businesses

AI Video Intelligence to Support Businesses Reopening

Today’s world has changed a lot since the start of the COVID19 pandemic. Health guidelines and business operational needs have…

Read Story

Never miss a minute

Get great content to your inbox every week. No spam.

    Only great content, we don’t share your email with third parties.