Abnormality detection within a defined area

A system for abnormality detection within a defined area, the system comprising:

• a server that includes a processor configured to:
• obtain a plurality of images of the defined area from one or more image-capture devices;
• compute, using a trained neural network, one or more types of products from an inference with each input image of the plurality of images, wherein the plurality of images are associated with a plurality of different locations in the defined area;
• determine, using the trained neural network, an extent of deviation properties of a probability distribution of each input image of the plurality of images with respect to a referential probability distribution of one or more reference images;
• compute, using the trained neural network, gradients associated with the trained neural network based on at least the determined extent of deviation for at least one input image of the plurality of images when the determined extent of deviation for the at least one input image is greater than a defined threshold;
• generate, using the trained neural network, a localized dimensional representation in a portion of the at least one input image associated with a first location of the plurality of locations, based on the computed gradients, wherein the generated localized dimensional representation provides a visual indication of an abnormality located in the first location within the defined area;
• and
• execute an action associated the first location in the defined area based on the generated localized dimensional representation, wherein the executed action is configured to cause a proactive control or prevention of an occurrence of an undesired event in the defined area.

A non-transitory computer-readable medium having stored thereon, computer implemented instructions that when executed by a computer causes the computer to execute operations, the operations comprising:

• obtaining a plurality of images of the defined physical area from one or more image-capture devices;
• computing, using a trained neural network, one or more types of products from an inference of each input image of the plurality of images, wherein the plurality of images are associated with a plurality of different locations in the defined area;
• determining, using the trained neural network, an extent of deviation of properties of a probability distribution of each input image of the plurality of images with respect to a referential probability distribution of one or more reference images;
• computing, using the trained neural network, gradients associated with the trained neural network based on at least the determined extent of deviation for at least one input image of the plurality of images when the determined extent of deviation for the at least one input image is greater than a defined threshold;
• generating, using the trained neural network, a localized dimensional representation in a portion of the at least one input image associated with a first location of the plurality of locations, based on the computed gradients, wherein the generated localized two-dimensional representation provides a visual indication of an abnormality located in the first location within the defined area;
• and
• executing an action associated the first location in the defined area based on the generated localized dimensional representation, wherein the executed action is configured to cause a proactive control or prevention of an occurrence of an undesired event in the defined area.