A method for recognizing handwritten text is disclosed. The method comprises receiving data comprising a sequence of ink points; applying the received data to a neural network-based sequence classifier trained with a Connectionist Temporal Classification (CTC) output layer using forced alignment to generate an output; generating a character hypothesis as a portion of the sequence of ink points; applying the character hypothesis to a character classifier to obtain a first probability corresponding to the probability that the character hypothesis includes the given character; processing the output of the CTC output layer to determine a second probability corresponding to the probability that the given character is observed within the character hypothesis; and combining the first probability and the second probability to obtain a combined probability corresponding to the probability that the character hypothesis includes the given character.

Systems and methods of image searching include receiving content, receiving a request to select an image from content, selecting a plurality of items in the image, retrieving information about the selected item, and providing display data based on the retrieved information.

Systems and methods of image searching include receiving content, receiving a request to select an image from content, selecting a plurality of items in the image, retrieving information about the selected item, and providing display data based on the retrieved information.

Aspects of the disclosure relate to training and using a phrase recognition model to identify phrases in images. As an example, a selected phrase list may include a plurality of phrases is received. Each phrase of the plurality of phrases includes text. An initial plurality of images may be received. A training image set may be selected from the initial plurality of images by identifying the phrase-containing images that include one or more phrases from the selected phrase list. Each given phrase-containing image of the training image set may be labeled with information identifying the one or more phrases from the selected phrase list included in the given phrase-containing images. The model may be trained based on the training image set such that the model is configured to, in response to receiving an input image, output data indicating whether a phrase of the plurality of phrases is included in the input image.

Systems and methods for digitized document image data spillage recovery are provided. One or more memories may be coupled to one or more processors, the one or more memories including instructions operable to be executed by the one or more processors. The one or more processors may be configured to capture an image; process the image through at least a first pass to generate a first contour; remove a preprinted bounding region of the first contour to retain text; generate one or more pixel blobs by applying one or more filters to smudge the text; identify the one or more pixel blobs that straddle one or more boundaries of the first contour; resize the first contour to enclose spillage of the one or more pixel blobs; overlay the text from the image within the resized contour; and apply pixel masking to the resized contour.

A method for estimating text heights of text line images includes estimating a text height with a sequence recognizer. The method further includes normalizing a vertical dimension and/or position of text within a text line image based on the text height. The method may also further include calculating a feature of the text line image. In some examples, the sequence recognizer estimates the text height with a machine learning model.

A smart reading device for a water meter and a controlling method thereof are provided. The smart reading device includes a fixing component, a casing, an image capturing component, an image analyzing component, and a transmitting component. The fixing component is used to be fixed onto the water meter. The casing is disposed on the fixing component. The image capturing component is disposed in the casing for capturing a numerical display area of the water meter so as to obtain a water consumption image. The water consumption image is analyzed by the image analyzing component or a relay device to obtain a water consumption value. The transmitting component is used for transmitting the water consumption value or the water consumption image to the relay device.

One or more highlights of a video stream may be identified. The highlights may be segments of a video stream, such as a broadcast of a sporting event, that are of particular interest to one or more users. According to one method, at least a portion of the video stream may be stored. The portion of the video stream may be compared with templates of a template database to identify the one or more highlights. Each highlight may be a subset of the video stream that is deemed likely to match the one or more templates. The highlights, an identifier that identifies each of the highlights within the video stream, and/or metadata pertaining particularly to the one or more highlights may be stored to facilitate playback of the highlights for the users.

Provided are a video title generation method, an electronic device and a storage medium, which relate to a technical field of video, and in particular to a technical field of short video. The method includes: obtaining a plurality of pieces of optional text information, for a first video file; determining central text information, from the plurality of pieces of optional text information, the central text information being optional text information with the highest similarity to content of the first video file; and determining the central text information as a title of the first video file. That is, an interest point in an original video file can be determined according to user's interactive behavior data on the original video file, and the original video file can be clipped based on the interest point to obtain a plurality of clipped video files, namely, short videos.

Provided are a video clipping and model training method, relating to the field of video technologies, and in particular, to the field of short video technologies. The video clipping method includes: acquiring interaction behavior data for an original video file; determining interaction heat at respective time points of the original video file, according to the interaction behavior data; selecting N time points with highest interaction heat, to take the selected time points as interest points of the original video file, where N is a positive integer; and clipping the original video file based on the respective interest points, to obtain N clipped video files. Therefore, high-quality short video files can be generated.