An example system includes a first and second digital device. The first digital device may be configured to provide an interface displaying an image including a depiction of an object, place a bounding shape around the object, and crop contents of the bounding shape to create a portion. The second digital device may be configured to receive the portion, retrieve high-level features and low-level features, apply first Atrous Spatial Pyramid Pooling (ASPP) to the high-level features to aggregate the high-level features as aggregate features, concatenate results to create the aggregate features, up-sample, apply a convolution to the low-level features, concatenate the aggregate features with the low-level features after convolution to form combined features, segment the combined features to generate a polygonal shape outline along outer boundaries of the first object, and provide the first polygonal shape outline to the first digital device for display.

A gas detection device, a gas detection method, and a gas detection program according to the present invention detect gas on the basis of an image obtained by imaging a detection target. The gas detection device, the gas detection method, and the gas detection program; generated accumulated data obtained by accumulating a number of times of gas detection in a predetermined unit of accumulation, on the basis of a plurality of images captured at a plurality of times different from each other in a predetermined periond; and generate a mask image for suppressing notification of detected gas on the basis of the generated accumulated data.

An image recognition device involves successively extracting co-occurrence pairs in synchronization with a clock, setting a weighting for the portion connecting the input layer and the intermediate layer corresponding to the extracted co-occurrence pairs, and successively inputting a first vote to the input layer. Meanwhile, the intermediate layer adds and stores the successively inputted number of votes. By continuing this operation, a value the same as if a histogram were inputted to an input layer is achieved in the intermediate layer, without creating a histogram. In this way, the image recognition device of this embodiment can perform image recognition while avoiding the creation of a histogram, which consumes vast amounts of memory. As a result of this configuration, it is possible to save memory resources, simplify circuits, and improve calculation speed, and achieve an integrated circuit suitable to an image recognition device.

The present disclosure provides an end-to-end attention pooling-based classification method for histopathological images. The method specifically includes the following steps: S1, cutting the histopathology image into patches of a specified size, removing the patches with too much background area and packaging the remaining patches into a bag; S2, training a deep learning network by taking the bag obtained in S1 as an input using a standard multi-instance learning method; S3, scoring all the patches by using the trained deep learning network, and selecting m patches with highest and lowest scores for each whole slide image to form a new bag; S4, building a deep learning network including an attention pooling module, and training the network by using the new bag obtained in S3; and S5, after the histopathology image to be classified is processed in S1 and S3, performing classification by using the model obtained in S4. The present disclosure can obtain a better classification effect under the current situation of only a small number of samples, provide an auxiliary diagnosis mechanism for doctors, and alleviate the problem of shortage of medical resources.

An aspect of the present invention allows for more accurately identifying a possible lesion in a human lung field. The aspect of the present invention includes an image obtaining section configured to obtain a chest cross-sectional image of a subject, a segmentation section configured to classify, into a plurality of segments, unit elements of the chest cross-sectional image, and an image dividing section configured to divide the chest cross-sectional image into a plurality of regions. A data deriving section is configured to derive data associated with the possible lesion, the data being derived on the basis of a segment of unit elements in the each region among the plurality of segments. An identifying section is configured to output an identification result, which is a result of identification of the possible lesion in the lung field of the subject.

A method includes: generating a refine image from an incorrect image from which an incorrect label is inferred by a neural network; generating a third map by superimposing a first map and a second map, the first map indicating pixels to each of which a change is made in generating the refine image, of plural pixels in the incorrect image, the second map indicating a degree of attention for each local region in the refine image, each local region being a region that has drawn attention at the time of inference by the neural network, and the third map indicating a degree of importance for each pixel for inferring a correct label; and obtaining an added value for respective divided region in the third map by summing pixel values within the respective divided region, the respective divided region being a region divided according to a predetermined index.

An image processing method includes: performing first segmentation processing on an image to be processed, and determining a segmentation region of a target in said image (S11); determining, according to the position of the center point of the segmentation region of the target, an image region where the target is located (S12); and performing second segmentation processing on the image region where each target is located, and determining the segmentation result of the target in said image (S13).

A weakly supervised image semantic segmentation method based on an intra-class discriminator includes: constructing two levels of intra-class discriminators for each image-level class to determine whether pixels belonging to the image class belong to a target foreground or a background, and using weakly supervised data for training; generating a pixel-level image class label based on the two levels of intra-class discriminators, and generating and outputting a semantic segmentation result; and further training an image semantic segmentation module or network by using the label to obtain a final semantic segmentation model for an unlabeled input image. By means of the new method, intra-class image information implied in a feature code is fully mined, foreground and background pixels are accurately distinguished, and performance of a weakly supervised semantic segmentation model is significantly improved under the condition of only relying on an image-level annotation.

The invention provides a method of generating an image recognition model and an electronic device using the method. The method includes the following. A source image is obtained; a first image is cut out of a first region of the source image to generate a cut source image; a preliminary image recognition model is pre-trained according to feature data and label data, in which the feature data is associated with the cut source image, and the label data is associated with the first image; and the pre-trained preliminary image recognition model is fine-tuned to generate the image recognition model. The method of generating the image recognition model and the electronic device provided by the invention may correctly restore an input image.

An image processing method includes displaying, by a terminal, a preview image, obtaining, by the terminal, at least one first image in the preview image, receiving, by the terminal, a photographing instruction, obtaining, by the terminal, at least one second image, detecting, by the terminal, a moving target based on the first image and the second image, and splitting, by the terminal, the moving target from the first image and the second image. The method further includes performing fusion processing on the first image and the second image after the splitting.