Abstract: Intelligent vehicle repair estimating techniques include an image processing component that extracts image attributes from one or more images of a damaged vehicle, and utilizes the attributes to predict an initial set of parts that are globally-identified. Based on a jurisdiction associated with the damaged vehicle, the initial set of parts is transformed into a set of jurisdictionally-based repairs (e.g., parts, labor operations, time intervals, costs, etc.), which may be included in a draft vehicle repair estimate. An estimate refinement component iteratively modifies/refines the draft estimate using a machine-only loop nested within a larger human-machine loop, where system-generated modifications are incrementally incorporated into the draft within the smaller loop, and user-generated modifications are incrementally incorporated into the draft within the larger loop.

Abstract: Techniques for automatic image tagging and selection at a mobile device include generating a smart image tagging model by first training an initial model based on different angles of image capture of subject vehicles, and then re-training the trained model using weights discovered from the first training and images that have been labeled with additional tags indicative of different vehicle portions and/or vehicle parameters. Nodes that are training-specific are removed from the re-trained model, and the lightweight model is serialized to generate the smart image tagging model. The generated model may autonomously execute at an imaging device to predict respective tags associated with a stream of frames; select, capture and store respective suitable frames as representative images corresponding to the predicted tags; and provide the set of representative images and associated tags for use in determining vehicle damage, insurance claims, and the like.

Abstract: Techniques for automatic image tagging and selection at a mobile device include generating a smart image tagging model by first training an initial model based on different angles of image capture of subject vehicles, and then re-training the trained model using weights discovered from the first training and images that have been labeled with additional tags indicative of different vehicle portions and/or vehicle parameters. Nodes that are training-specific are removed from the re-trained model, and the lightweight model is serialized to generate the smart image tagging model. The generated model may autonomously execute at an imaging device to predict respective tags associated with a stream of frames; select, capture and store respective suitable frames as representative images corresponding to the predicted tags; and provide the set of representative images and associated tags for use in determining vehicle damage, insurance claims, and the like.

Abstract: Techniques for optimizing vehicle license plate recognition in images and their decoding include training a set of convolutional neural networks (CNNs) by using images in which license plates are identified or labeled as a whole, rather than by license plate parts or key points, and rather than by the individual, segmented characters represented thereon. The trained CNNs may operate on target images of environments to localize images of license plates included therein and determine the issuing jurisdiction and/or ordered set of characters represented on detected license plates without utilizing character segmentation and/or per-character recognition techniques. As such, license plates depicted within target images are able to be detected and decoded with greater tolerances for lighting conditions, deformations or damages, occlusions, differing image resolutions, differing angles of capture, variations of other objects depicted within the images (such as dense or changing signage), etc.

Abstract: Intelligent vehicle repair estimating techniques include an image processing component that extracts image attributes from one or more images of a damaged vehicle, and utilizes the attributes to predict an initial set of parts that are globally-identified. Based on a jurisdiction associated with the damaged vehicle, the initial set of parts is transformed into a set of jurisdictionally-based repairs (e.g., parts, labor operations, time intervals, costs, etc.), which may be included in a draft vehicle repair estimate. An estimate refinement component iteratively modifies/refines the draft estimate using a machine-only loop nested within a larger human-machine loop, where system-generated modifications are incrementally incorporated into the draft within the smaller loop, and user-generated modifications are incrementally incorporated into the draft within the larger loop.