Glossary

Overall, how often is the algorithm correct? Not as good of a measure as F1, but it is often colloquially used to describe algorithm performance.

(Artificial Intelligence) the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.

In AI/ML, a model replicates a decision process to enable automation and understanding. AI/ML models are mathematical algorithms that are “trained” using data and human expert input to replicate a decision an expert would make when provided that same information

Add media to the application by uploading it to an input subject or capturing it from a live stream. Provide feedback on the media to train the application to automatically detect, measure, or count the items of interest.

An application is a unit of work that performs a single visual task. The Cogniac system supports various application types, including classification, detection, measurement, localization, and counting. Many of these applications are based on deep convolutional neural network models that have been trained to perform specific tasks.

Also known in the industry as semantic segmentation. A type of model that, given an image, returns the exact pixels that are associated with that object. Example, an area detection model that is trained to identify a crack in glass, will return the exactly pixels that represent that crack.

The Meraki dashboard API is a software interface that interacts directly with the Meraki cloud platform and Meraki-managed devices.

A type of model that, given an image, produces a exactly one result that represents the entire image. Example, a classifier that is trained with output of 'cat' and 'dog'. Given a picture of a cat, result is 'cat'. Given a picture of a dog, result is 'dog'. Given a picture of a flower, will still result in a prediction of either 'cat' or 'dog'. 

Cogniac’s CloudCore is the centralized workflow development, model training, and evaluation platform. It is a powerful, centralized hub for interacting with your entire solution.

Convolutional Neural Network - It is a type of deep neural network often used for computer vision tasks. Due to the effectiveness of CNNs, they have numerous applications in facial recognition, medical image analysis, and natural language processing tasks.

(Collaborative Robot ) Cobots, or collaborative robots, are robots intended for direct human robot interaction within a shared space, or where humans and robots are in close proximity. Cobot applications contrast with traditional industrial robot applications in which robots are isolated from human contact.

Value produced by a model representing the probability that a given detection is correct.

(Central Processing Unit) A CPU works together with a GPU to increase the throughput of data and the number of concurrent calculations within an application. ... Using the power of parallelism, a GPU can complete more work in the same amount of time as compared to a CPU

Media flows through the Cogniac system via subjects and applications. In the simplest case, media is uploaded into an input subject and processed by an application. The application then associates the media with an output subject, which can be used as input for another application. This allows for complex processing pipelines to be created.

Cogniac takes data security seriously and follows industry best practices to protect your data. We use a private tenant architecture model, meaning each customer's data is siloed and isolated from other customers. This ensures that only the users you invite to your tenant can access your media, subjects, applications, and associated data.

(DL) Deep learning is part of a broader family of machine learning methods based on artificial neural networks with representation learning. Learning can be supervised, semi-supervised or unsupervised.

Machine learning uses algorithms to parse data, learn from that data, and make informed decisions based on what it has learned. ... Deep learning is a subfield of machine learning. While both fall under the broad category of artificial intelligence, deep learning is what powers the most human-like artificial intelligence

The application generates detections as media is uploaded or captured into the input subjects. These detections can be automated or human-generated, created from items of interest within the user's visual media.

Detection Full Frame

The harmonic mean of Precision and Recall. This is often a more fair look at the overall performance of the algorithm than accuracy. 

(False Negative) The algorithm predicted false and the ground truth is true. Example: Algorithm predicts that this image does not contain the defect of interest and the image actually does contain the defect of interest.

(False Positive) The algorithm predicted true but the ground truth is false. Example: Algorithm predicts that this image contains the defect of interest but the image does not actually contains the defect of interest.

Gigabit Ethernet (GigE) Is the way we talk about the technology for transmitting data at a rate of a gigabit per second (or 1,000 megabits per second). That's ten times faster than the other three ports, which transmit data at 100 megabits per second.

Graphics Processing Unit

In machine learning, the term "ground truth" refers to the accuracy of the training set classification for supervised learning techniques. This is used in statistical models to prove or disprove research hypotheses.

(HPO) In machine learning, a hyperparameter is a parameter whose value is used to control the learning process. By contrast, the values of other parameters are derived via training.

Inference refers to the process of using a trained machine-learning algorithm to make a prediction. Inference can be performed on a supported Meraki MV camera.

Collections of related media that are meant to be processed by the Application Model.

A knowledge base is a self-serve online library of information about a product, service, department, or topic. The data in your knowledge base can come from anywhere. Typically, contributors who are well versed in the relevant subjects add to and expand the knowledge base. The content can range from the ins and outs of your HR or legal department to an explanation of how a product works. 

The process of providing ground truth for a set of data(identifying the areas of interest), i.e. labeling images. Most commonly done by a human. 

Machine learning is a subfield of artificial intelligence, which is broadly defined as the capability of a machine to imitate intelligent human behavior. Artificial intelligence systems are used to perform complex tasks in a way that is similar to how humans solve problems.

(Machine Learning) the use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyze and draw inferences from patterns in data.

The Cogniac platform trains an algorithm with parameters (i.e., a Model) that allows the system to infer if an image includes the target that interests the user. The target could be the presence of a defect or a city bus, depending on the use case. This model will be uploaded to the Meraki MV camera, allowing it to make inferences on images and communicate them to the Meraki Dashboard.

(Neural Network) A neural network is either a system software or hardware that works similar to the tasks performed by neurons of human brain. Neural networks include various technologies like deep learning, and machine learning as a part of Artificial Intelligence

(Optical Character Recognition) Optical character recognition or optical character reader is the electronic or mechanical conversion of images of typed, handwritten or printed text into machine-encoded text, whether from a scanned document, a photo of a document, a scene-photo or from subtitle text superimposed on an image

 The items or conditions you plan to detect from within images or video.

When the algorithm predicts yes, how often is it correct? In other words, of the items that the algorithm found to be true, how many actually are true. You would want to optimize for precision in cases when having too many false detections is costly

Users create tenants and virtual workspaces for storing and managing media and applications. Users can then invite other users to join the tenant. Once the tenant is created, users can create their first application.

When it's actually 'yes', how often does the algorithm predict 'yes' оf the items that are true in the ground truth, how many did we find as true. You would want to optimize for recall in cases when missing a true detection is costly

(Real Time Streaming Protocol) The Real Time Streaming Protocol (RTSP) is a network control protocol designed for use in entertainment and communications systems to control streaming media servers. The protocol is used for establishing and controlling media sessions between endpoints

A subject is the central means by which concepts can be associated with media in the Cogniac system. Subjects also provide a mechanism by which media can be grouped and managed within the Cogniac system

A subject is a user-defined concept that is used to group, manage, and route media within the Cogniac system. Subjects are typically related to the goals of the visual observation task that is being automated. For example, a subject could be "defective gears" or "cats with mouse in mouth."

An algorithm uses training data and feedback from humans to learn the relationship of given inputs to a given output (eg, how the inputs “time of year” and “interest rates” predict housing prices) 

A tenant in the CloudCore system is a dedicated, isolated workspace for a particular customer, use case or end-user application. Tenants are securely isolated from each other in the CloudCore system, with each tenant having independent user lists and flexible Role-Based Access Controls.

(True Negative) The algorithm predicted false and the ground truth is false. Example: Algorithm predicts that this image does not contain the defect of interest and the image actually does not contain the defect of interest.

(True Positive) The algorithm predicted true and the ground truth is also true. Example: Algorithm predicts that this image contains the defect of interest and the image actually contains the defect of interest.