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: Depth information for each pixel of a digitally captured image of a document occluded by an object is determined. The pixels the pixels for which the depth information is indicative of a greater depth than a baseline depth of the document are identified as belonging to an artifact corresponding to the object occluding the document. The artifact is removed from the digitally captured image.

Abstract: According to examples, an apparatus may include a processor and a memory on which are stored machine-readable instructions that when executed by the processor, may cause the processor to receive an image and identify contents in the received image. The processor may identify candidate regions on the image at which a predefined object is placeable. In some examples, the processor may assign scores to the identified candidate regions based on relative positions of the identified candidate regions to respective ones of the identified contents in the image. Based on the assigned scores, the processor may select a candidate region among the identified candidate regions at which the predefined object is to be placed. The processor may determine a size and a position of the predefined object based on the selected candidate region, and may output the determined size and the position of the predefined object on the image.

Abstract: Upon placement of a camera-facing surface of a camera device on a document or upon parallel positioning of the camera-facing surface close to and over the document, images are continually captured by an image capturing sensor of the camera device. While the camera device is being raised above the document, whether the document is fully included within a captured image is detected. In response to detecting that the document is fully included within the captured image, the captured image that fully includes the document is selected as a document image.

Abstract: A contextual feature matrix that aggregates contextual information within a captured image of a document at multiple scales is generated using a multiscale aggregator machine learning model. Pixel-wise enhancement curves for the captured image are estimated based on the contextual feature matrix using an enhancement curve prediction machine learning model. The pixel-wise enhancement curves are iteratively applied to the captured image to enhance the document within the captured image.

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.