Systems and methods described herein utilize synthetic pixel generation using a custom neural network to generate synthetic versions of objects hidden by occlusions for effective detection and tracking. A computing device stores an object detector model and a synthetic image generator model; receives a video feed; detects objects of interest in a current frame of the video feed; identifies an occluded object in the current frame; retrieves a previous frame from the video feed; generates synthetic data based on the previous frame for the occluded object; and forwards a modified version of the current frame to an object tracking system, wherein the modified version of the current frame includes the synthetic data.

Disclosed is an approach of on-device partial recognition that includes performing partial recognition on an image of a document captured by a mobile device to detect and/or recognize a specific area (e.g., barcodes, non-relevant text, etc.) and filling the recognized area with a solid color. Because the solid color area has a maximum compression ratio, this approach can lead to image size reduction and increased network throughput for client-server based data recognition where further processing such as advanced data extraction is performed at the server side. The approach can be enforced with neural network algorithms to exclude non-relevant information (e.g., logos, phrases, words, etc.).

An identification apparatus includes processing circuitry configured to determine whether or not a first image and a second image are similar based on feature points extracted from each of the first image and the second image, and determine whether or not the first image and the second image are similar by comparing neighborhood graphs generated for each of the first image and the second image, the feature points being as nodes.

Methods and systems are provided for optimal segmentation of an image based on multiple segmentations. In particular, multiple segmentation methods can be combined by taking into account previous segmentations. For instance, an optimal segmentation can be generated by iteratively integrating a previous segmentation (e.g., using an image segmentation method) with a current segmentation (e.g., using the same or different image segmentation method). To allow for optimal segmentation of an image based on multiple segmentations, one or more neural networks can be used. For instance, a convolutional RNN can be used to maintain information related to one or more previous segmentations when transitioning from one segmentation method to the next. The convolutional RNN can combine the previous segmentation(s) with the current segmentation without requiring any information about the image segmentation method(s) used to generate the segmentations.

The present disclosure relates to a dimension measuring device that shortens a time required for dimension measurement and eliminates errors caused by an operator. A dimension measuring device that measures a dimension of a measurement target using an input image is provided, in which a first image in which each region of the input image is labeled by region is generated by machine learning, an intermediate image including a marker indicating each region of the first image is generated based on the generated first image, a second image in which each region of the input image is labeled by region is generated based on the input image and the generated intermediate image, coordinates of a boundary line between adjacent regions are obtained by using the generated second image, coordinates of a feature point that defines a dimension condition of the measurement target are obtained by using the obtained coordinates of the boundary line, and the dimension of the measurement target is measured by using the obtained coordinates of the feature point.

2D and 3D data of a scene are linked by associating points in the 3D data with corresponding points in multiple different 2D images within the 2D data. Labels assigned to points in either data can be propagated to the other data. Labels propagated to a point in the 3D data are aggregated, and the labels ranked highest are kept and propagated back to the 2D images. 3D data including labels produced in this manner allow partially obscured objects in certain views to be more accurately identified. Thus, an object can be manipulated in all 2D views of the 2D data in which the object is at least partially visible, in order to digitally remove, alter, or replace it.

A method and computer programs for detecting document regions that will be excluded from a watermark embedding process are disclosed. The method comprises converting, by an adapter module, at least one page of a received document into a visual representation thereof, the visual representation keeping the position of the characters of the at least one page; receiving, by a text detector, the visual representation; processing, by the text detector, the visual representation using one or more artificial intelligence algorithms, and returning a list of invalid regions with their associated page positions as a result, wherein each invalid region of the list of invalid regions may have associated thereto a confidence score; and using, by a watermark embedding module or by a watermark extracting module, the list of invalid regions to provide a watermarked document or a message embedded in the document.

Methods, systems, and non-transitory computer readable media are disclosed for accurately and efficiently generating groups of images portraying semantically similar objects for utilization in building machine learning models. In particular, the disclosed system utilizes metadata and spatial statistics to extract semantically similar objects from a repository of digital images. In some embodiments, the disclosed system generates color embeddings and content embeddings for the identified objects. The disclosed system can further group similar objects together within a query space by utilizing a clustering algorithm to create object clusters and then refining and combining the object clusters within the query space. In some embodiments, the disclosed system utilizes one or more of the object clusters to build a machine learning model.

Disclosed is a system and a method for monitoring a CT scan image. A CT scan image may be resampled into a plurality of slices using a bilinear interpolation. A region of interest may be identified on each slice using an image processing technique. The region of interest may be masked on each slice using deep learning. Subsequently, a nodule may be detected as the region of interest using the deep learning. Further, a plurality of characteristics associated with the nodule may be identified. Furthermore, an emphysema may be detected in the region of interest on each slice. A malignancy risk score for the patient may be computed. A progress of the nodule may be monitored across subsequent CT scan images. Finally, a report of the patient may be generated.

A label-free cell classification and screening system based on hybrid transfer learning, including a data preprocessing module for acquiring 2D light scattering video data and for digital cell filtering, is made public here; the data preprocessing module includes the label-free high-content video flow cytometry, which has the optical excitation module, the sheath flow control module, and the data acquisition and processing module; the image archiving module is used to sort and set labels for cells; in the feature extraction module, the first convolutional neural network is used to obtain image data feature vectors; in the cell classification and screening module, a support vector machine model is used to obtain the cell screening results.