Abstract: A backbone machine learning model is applied to a captured image of a number of documents to identify image features of the captured image. A head machine learning model is applied to the image features to identify a number of candidate quadrilaterals within the captured image. An edge detection technique is applied to each candidate quadrilateral to identify edges of a corresponding document of the captured image within the candidate quadrilateral.

Abstract: A device including a camera, a user interface, a memory, and at least one processor. The camera is to capture a plurality of images. The user interface is to display a captured image among the plurality of captured images and to display feedback corresponding to the captured image. The memory is to store feedback data predicting a quality threshold of the plurality of captured images and indicating a type of feedback to be displayed on the user interface corresponding to the captured image. The at least one processor is to display the feedback on the user interface based on whether the captured image is below the predicted quality threshold of the plurality of captured images, the quality threshold being based on the stored feedback data, and update the feedback data stored in the memory based on a comparison between the captured image and the plurality of captured images.

Abstract: A point extraction machine learning model is applied to a captured image of one or multiple documents to identify the documents within the captured image and to identify boundary points for each document. For each document identified within the captured image, an instance segmentation machine learning model is applied to the boundary points for the document and to the captured image to extract a segmentation mask for the document.

Abstract: A method of protecting privacy in video may include classifying, via a trained multi-level neural network, a plurality of objects within frames of a video content. Training the multi-level neural network may include inputting a plurality of pixels of at least one frame, fitting the pixels into different layers, and assigning weights to individual components within the layers. The method may also include determining if the classified objects are private objects or non-private objects, tracking the detected objects between each frame of the video content within a threshold distance of movement, and masking the private objects.

Abstract: An example non-transitory computer-readable medium includes instructions executable by a processor to detect boundaries of a representation of a document page in a captured image, model the boundaries of the representation of the document page as nonlinear curves, use the nonlinear curves to transform pixels of the representation of the document page into pixels of a dewarped representation of the document page, and output a dewarped image based on the dewarped representation of the document page.

Abstract: A method of protecting privacy in video may include classifying, via a trained multi-level neural network, a plurality of objects within frames of a video content. Training the multi-level neural network may include inputting a plurality of pixels of at least one frame, fitting the pixels into different layers, and assigning weights to individual components within the layers. The method may also include determining if the classified objects are private objects or non-private objects, tracking the detected objects between each frame of the video content within a threshold distance of movement, and masking the private objects.

Abstract: In example implementations, an apparatus is provided. The apparatus includes a video camera, a detection device, a tracking device, a display, and a processor. The video camera is to capture a video image of a plurality of documents. The detection device is to detect the plurality of documents in a frame of the video image. The tracking device is to track each one of the plurality of documents that is detected in a series of frames of the video image. The display is to generate a visual indicator around each one of the plurality of documents in the video image that is displayed. The processor is in communication with the video camera, the detection device, the tracking device, and the display to control execution of the video camera, the detection device, the tracking device, and the display.

Abstract: Examples of methods performed by an electronic device are described. In some examples of the methods, a machine learning model is trained based on a plurality of interaction events and a corresponding plurality of images. In an example, each of the plurality of interaction events corresponds to one of a plurality of displays. In some examples of the methods, a display is selected of the plurality of displays based on the machine learning model. In an example, an object is presented on the display.

Abstract: Example implementations relate to hiding sensitive data. For example, a device according to the present disclosure may include a scanning portion to receive a scan of a document, and a graphical user interface to receive a command with respect to the sensitive data. The device may also include a processor to analyze the scan of the document, process the command via the graphical user interface, determine the document contains sensitive data in response to the command, and perform an action to hide the sensitive data.

Abstract: Example implementations relate to hiding sensitive data. For example, a device according to the present disclosure may include a scanning portion to receive a scan of a document, and a graphical user interface to receive a command with respect to the sensitive data. The device may also include a processor to analyze the scan of the document, process the command via the graphical user interface, determine the document contains sensitive data in response to the command, and perform an action to hide the sensitive data.