Broadly speaking, the present techniques generally relate to a method and apparatus for video recognition, and in particular relate to a computer-implemented method for performing video recognition using a transformer-based machine learning, ML, model. Put another way, the present techniques provide new methods of image processing in order to automatically extract feature information from a video.

A feature extractor and novel training objective are provided for content-based image retrieval. For example, a computer-implemented method includes applying a query image and a search image to a neural network of a feature extraction network of a computing device, the query image indicating an object to be searched for in the search image. The feature extraction network includes the neural network, a spatial feature neural network receiving a first output of the neural network pertaining to the search image, and an embedding network receiving a second output of the neural network pertaining to the query image. The method includes generating spatial search features from the spatial feature neural network, generating a query feature from the embedding network, applying the query feature to an artificial neural network (ANN) index, and determining an optimal matching result of an object in the search image based on an operation using the ANN index.

A feature extractor and novel training objective are provided for content-based image retrieval. For example, a computer-implemented method includes applying a query image and a search image to a neural network of a feature extraction network of a computing device, the query image indicating an object to be searched for in the search image. The feature extraction network includes the neural network, a spatial feature neural network receiving a first output of the neural network pertaining to the search image, and an embedding network receiving a second output of the neural network pertaining to the query image. The method includes generating spatial search features from the spatial feature neural network, generating a query feature from the embedding network, applying the query feature to an artificial neural network (ANN) index, and determining an optimal matching result of an object in the search image based on an operation using the ANN index.

A defect inspecting system includes a detector configured to image a sample and a host control device that acquires an inspection image including a defect and a plurality of reference images not including a defect site and generates a pseudo defect image by editing a predetermined reference image among the plurality of acquired reference images. An initial parameter is determined with which the pseudo defect site is detectable from the pseudo defect image. The host control device acquires a defect candidate site from the inspection image using the initial parameter, estimates a high-quality image from an image of a site corresponding to the defect candidate site using the parameter acquired in image quality enhancement, and specifies an actual defect site in the inspection image by executing defect discrimination. A parameter is determined with which a site close to the specified actual defect site is detectable using the inspection image.

The invention relates to the field of computer engineering for processing images that provides increased accuracy of finding and classifying a similar object . The technical result is achieved by: downloading files of a radiographic image which comprise metadata including information about the object or subject of the image and information about the image itself; encrypting the downloaded files if the above-mentioned files comprise personal data about a person; decrypting the above-mentioned, encrypted, downloaded files; and processing the radiographic image, wherein, as a result of the processing, the following occurs: finding and capturing a relevant region of the radiographic image; removing noise from the captured, relevant region of the radiographic image, wherein a region with a found object is meant by a relevant region of the radiographic image; compressing or unzipping a previously processed radiographic image; and finding a similar object in two previously processed images, and processing said object.

The invention relates to the field of computer engineering for processing images that provides increased accuracy of finding and classifying a similar object . The technical result is achieved by: downloading files of a radiographic image which comprise metadata including information about the object or subject of the image and information about the image itself; encrypting the downloaded files if the above-mentioned files comprise personal data about a person; decrypting the above-mentioned, encrypted, downloaded files; and processing the radiographic image, wherein, as a result of the processing, the following occurs: finding and capturing a relevant region of the radiographic image; removing noise from the captured, relevant region of the radiographic image, wherein a region with a found object is meant by a relevant region of the radiographic image; compressing or unzipping a previously processed radiographic image; and finding a similar object in two previously processed images, and processing said object.

Methods and apparatuses for training a magnetic resonance imaging model, electronic devices and computer readable storage media are provided. A method may include: acquiring a magnetic resonance image data set; constructing a ring deep neural network to be trained; inputting an under-sampled magnetic resonance image and a full-sampled magnetic resonance image respectively to two neural networks included in the ring deep neural network, to generate respective simulated magnetic resonance images; inputting a first simulated full-sampled magnetic resonance image and the full-sampled magnetic resonance image to a pre-constructed first simulated magnetic resonance image class discrimination model, to obtain a first discrimination result indicating whether or not the first simulated full-sampled magnetic resonance image is of a simulated magnetic resonance image class; and adjusting a network parameter of the ring deep neural network based on a preset loss function, to obtain a trained magnetic resonance imaging model.

The invention relates to a computer-implemented method comprising:—acquiring (406) first training images (108) using a first image acquisition technique (104), each first training image depicting a plant-related motive; —acquiring (402) second training images (106) using a second image acquisition technique (102), each second training image depicting the motive depicted in a respective one of the first training images; —automatically assigning (404) at least one label (150, 152, 154) to each of the acquired second training images; —spatially aligning (408) the first and second training images which are depicting the same one of the motives into an aligned training image pair; —training (410) a machine-learning model (132) as a function of the aligned training image pairs and the labels, wherein during the training the machine-learning model (132) learns to automatically assign one or more labels (250, 252, 254) to any test image (205) acquired with the first image acquisition technique which depicts a plant-related motive; and—providing (412) the trained machine-learning model (132).

The disclosure relates to a method for making sensor data more robust to adversarial perturbations, wherein sensor data are obtained from at least two sensors, wherein the sensor data obtained from the at least two sensors are replaced in each case piecewise by means of quilting, wherein the piecewise replacement is carried out in such a way that the respectively replaced sensor data from different sensors are plausible relative to one another, and wherein the sensor data replaced piecewise are output.

The invention relates a computer-implemented method (500) of generating explainability information for explaining a model output of a trained model. The method uses one or more aspect recognition models configured to indicate a presence of respective characteristics in the input instance. A saliency method is applied to obtain a masked source representation of the input instance at a source layer of the trained model (e.g., the input layer or an internal layer), comprising those elements at the source layer relevant to the model output. The masked source representation is mapped to a target layer (e.g., input or internal layer) of an aspect recognition model, and the aspect recognition model is then applied to obtain a model output indicating a presence of the given characteristic relevant to the model output of the trained model. As explainability information, the characteristics indicated by the aspect recognition models are output.