Abstract: Disclosed herein are methods and systems for training machine learning (ML) models to classify activity of objects, comprising selecting a set of frames depicting one or more objects from one or more video sequences each comprising a plurality of consecutive frames, associating the object(s) with each pixel included in a bounding box of the object(s) identified in each frame of the set, computing a motion mask for each frame of the set indicating whether each pixel associated with the object(s) in the frame is changed or unchanged compared to a corresponding pixel in a preceding frame, augmenting an image of the object(s) in each frame of a subset of frames of the set to depict only the changed pixels associated with the object(s) by cutting out the unchanged pixels, and training one or more ML models, using the set of frames, to classify one or more activities of the object(s).

Abstract: A computer implemented method of identifying a behavior of objects detected in a video stream, comprising using one or more processors adapted for receiving a video stream comprising a plurality of frames captured by imaging sensor(s), applying a first set of trained classification functions for visually analyzing each of at least a subset of the frames to create a semantic attribute set for each object detected in each frame comprising an object probability score indicative of the object class probability, attribute probability score(s) indicative of probability of attribute(s) associated with the object and estimated location of the object, applying a second set of trained classification functions for semantically analyzing the attribute sets created for each object tracked in the subset to calculate a behavior probability score indicative of action(s) estimated to be conducted by each object and outputting an indication when the calculated behavior probability score complies with predefined rule(s).

Abstract: A computer implemented method of identifying a behavior of objects detected in a video stream, comprising using one or more processors adapted for receiving a video stream comprising a plurality of frames captured by imaging sensor(s), applying a first set of trained classification functions for visually analyzing each of at least a subset of the frames to create a semantic attribute set for each object detected in each frame comprising an object probability score indicative of the object class probability, attribute probability score(s) indicative of probability of attribute(s) associated with the object and estimated location of the object, applying a second set of trained classification functions for semantically analyzing the attribute sets created for each object tracked in the subset to calculate a behavior probability score indicative of action(s) estimated to be conducted by each object and outputting an indication when the calculated behavior probability score complies with predefined rule(s).

Abstract: A computer implemented method of semantically categorizing a video stream through multimodal content classification, comprising dividing a designated video stream to a plurality of scenes by analyzing a visual content of a plurality of frames of the video stream to identify scene changes between consecutive scenes, applying a plurality of classification functions to each of a plurality of modalities extracted from each of the scenes to calculate a class probability for each of a plurality of known concepts detected in each scene, applying a plurality of multimodal classification functions on the class probability of the known concepts to calculate a scene category probability for each scene indicating a probability of the scene to be categorized in one or more semantic categories and categorizing the video stream to a stream category of the semantic categories by aggregating the category probability of the scenes.

Abstract: A computer implemented method of detecting scene changes in a stream of frames by analyzing content differences between consecutive frames, comprising: (a) Identifying a content of each of a plurality of frames of a frames stream by applying a plurality of visual classification functions to each frame. The content comprises one or more of a plurality of visual elements. (b) Determining a content difference between every two consecutive frames of the plurality of frames by comparing the content of the two consecutive frames. (c) Detecting a scene change between the two consecutive frames when the content difference exceeds a pre-defined threshold. The scene change defines a separation between consecutive scenes of a plurality of scenes in the frames stream wherein each of the scenes comprises a subset of the plurality of frames.

Abstract: A computer implemented method of semantically categorizing a video stream through multimodal content classification, comprising dividing a designated video stream to a plurality of scenes by analyzing a visual content of a plurality of frames of the video stream to identify scene changes between consecutive scenes, applying a plurality of classification functions to each of a plurality of modalities extracted from each of the scenes to calculate a class probability for each of a plurality of known concepts detected in each scene, applying a plurality of multimodal classification functions on the class probability of the known concepts to calculate a scene category probability for each scene indicating a probability of the scene to be categorized in one or more semantic categories and categorizing the video stream to a stream category of the semantic categories by aggregating the category probability of the scenes.

Abstract: A computer implemented method of detecting scene changes in a stream of frames by analyzing content differences between consecutive frames, comprising: (a) Identifying a content of each of a plurality of frames of a frames stream by applying a plurality of visual classification functions to each frame. The content comprises one or more of a plurality of visual elements. (b) Determining a content difference between every two consecutive frames of the plurality of frames by comparing the content of the two consecutive frames. (c) Detecting a scene change between the two consecutive frames when the content difference exceeds a pre-defined threshold. The scene change defines a separation between consecutive scenes of a plurality of scenes in the frames stream wherein each of the scenes comprises a subset of the plurality of frames.