There is provided a multi-scale object detection device. The device includes an image frame acquisition unit for acquiring a plurality of consecutive image frames, a critical region extractor for extracting at least one second critical region from a current image frame based on at least one first critical region extracted from a previous image frame among the consecutive image frames, a multi-scale object detector whose operation involves a first object detection process for the current image frame and a second object detection process for the at least one second critical region, and an object detection integration unit for integrating the results of the first and second object detection processes.

A gas detection device that gives a notification of a detected gas on the basis of a captured image obtained by capturing an image of a monitoring target, the gas detection device including: a gas detection unit that detects gas on the basis of the captured image and gives a notification of the detected gas; an input unit that receives input information from a user; a mask candidate region extraction unit that extracts a mask candidate region that is a candidate region of a mask region for which a notification of gas detection is suppressed; and a mask generation unit that generates mask data indicating the mask region, in which the gas detection unit gives a notification of a gas detected outside the mask region, and the mask generation unit generates, as the mask data, a region in which first mask candidate region information input from the input unit matches second mask candidate region information extracted by the mask candidate region extraction unit.

A computing system for detecting objects in an image can perform operations including generating an image pyramid that includes a first level corresponding with the image at a first resolution and a second level corresponding with the image at a second resolution. The operations can include tiling the first level and the second level by dividing the first level into a first plurality of tiles and the second level into a second plurality of tiles; inputting the first plurality of tiles and the second plurality of tiles into a machine-learned object detection model; receiving, as an output of the machine-learned object detection model, object detection data that includes bounding boxes respectively defined with respect to individual ones of the first plurality of tiles and the second plurality of tiles; and generating image object detection output by mapping the object detection data onto an image space of the image.

A method for validating images with observations is disclosed. The method includes receiving production image and baseline image from user device. The method includes identifying a plurality of deviations in the production image from the baseline image using a similarity check algorithm. The method includes extracting first output corresponding to the plurality of deviations. The method includes creating, using cognizance Region of Interest (ROI) algorithm, one or more ROIs based on the position coordinates of each of the plurality of deviations. Each of the one or more ROIs comprises at least one deviation of the plurality of deviations. The cognizance ROI algorithm is based on Artificial Intelligence (AI). The method includes, for each ROI, generating, a second output corresponding to the ROI using one or more predictive models. The second output includes observations corresponding to each of the at least one deviation in the ROI.

Disclosed herein is a method for detecting an anomaly state based on screen output. The method includes receiving the output screen of a target device to be monitored, setting a target region to be examined in the output screen of the target device to be monitored, calculating a feature value vector corresponding to the state of the target region to be examined, calculating an anomaly score using a pretrained auto-encoder by receiving the feature value vector as input, and determining whether the target device to be monitored is anomalous using the anomaly score.

A method, apparatus, and computer program product for face liveness detection are disclosed. The method comprises: obtaining one or more color image data frames, each color image data frame depicting a face of a subject; identifying a plurality of skin regions; extracting a skin region data set from each one of the plurality of identified skin regions; computing a plurality of color distributions, each color distribution being computed on the basis of one of the plurality of skin region data sets; determining at least one distance between the plurality of color distributions; if the at least one distance is greater than a liveness threshold, detecting positive liveness of the subject, and else detecting negative liveness of the subject; and outputting the detected positive or negative liveness.

In some implementations, a device may receive an image of a document, the image depicting a reference feature associated with the document, the reference feature including at least one of: a face of a person, a machine-readable code, or a text field. The device may identify a rotational angle of the reference feature as depicted in the image based on comparing the reference feature as depicted in the image to one or more orientation parameters of the reference feature associated with a display orientation associated with the document. The device may rotate the image of the document by an angle to obtain an orientated image of the document, the angle being based on the rotational angle of the reference feature as depicted in the image. The device may provide the orientated image of the document for display.

Detecting fine-grained similarity in image includes determining a core area of a search image by generating an image salient map from a plurality of layers of the search image and determining a connected area based on the image salient map. Feature descriptors are generated from the core area of the search image. A plurality of capsule vectors are generated from different ones of a plurality of keypoints of the feature descriptors. Capsule vectors of the search image are compared with capsule vectors of each image of the dataset to generate a top-K matrix. Similarity scores for the top-K matrix are calculated. One or more image of the dataset having fine-grained similarity with the search image are selected based a bundled similarity score for each image of the dataset. The bundled similarity score is a summation of the similarity scores of the image.

Systems and methods for image segmentation are described. Embodiments of the present disclosure receive a training image and a caption for the training image, wherein the caption includes text describing an object in the training image; generate a pseudo mask for the object using a teacher network based on the text describing the object; generate a mask for the object using a student network; compute noise information for the training image using a noise estimation network; and update parameters of the student network based on the mask, the pseudo mask, and the noise information.

A method for calibrating a digital image acquisition system includes acquiring a digital image of a calibration target. Locations of each of a plurality of identifying features in the calibration target are determined and distances are computed between selected ones of the plurality of identifying features. A calibration grid is computed and overlay ed on the acquired digital image. The calibration grid is computed from a location of a reference one of the plurality of identifying features in the acquired digital image, the computed distances between the selected ones of the plurality of identifying features, and known locations of the plurality of calibration regions with respect to the reference one of the plurality of identifying features in the calibration target. The calibration grid specifies a plurality of calibration areas that correspond to the plurality of calibration regions in the calibration target.