A device for classifying image data includes a trainable pre-processing unit configured to retrieve, from a trained context, and based on the image data, at least one specification in terms of how a degree of granulation of the image data is to be reduced, and to reduce the degree of granulation of the image data in accordance with the at least one specification. The device further includes a trainable classifier configured to map the granulation-reduced image data onto an assignment to one or more classes of a specified classification.

An obstacle detection method can improve the accuracy of determining a relative positional relationship between two or more obstacles that are obstructed or obscured during automated driving. A road scene image of a road where a target vehicle is located is acquired. Obstacle recognition is performed to obtain region information and depth-of-field information corresponding to each obstacle in the road scene image. Target obstacles in an occlusion relationship and a relative depth-of-field relationship between the target obstacles are determined. A ranging result of each obstacle is acquired using a ranging apparatus corresponding to the target vehicle. An obstacle detection result of the road is determined based on the relative depth of field relationship between the target obstacles and the ranging result of each obstacle, thereby improving the accuracy of determining a positional relationship of obstructed or obscured obstacles during automated driving.

A system and a method for super-resolution image processing in remote sensing are disclosed. One or more sets of multi-temporal images with an input resolution and one or more first target images with a first output resolution are generated from one or more data sources. The first output resolution is higher than the input resolution. Each set of multi-temporal images is processed to improve an image match in the corresponding set of multi-temporal images. The one or more sets of multi-temporal images are associated with the one or more first target images to generate a training dataset. A deep learning model is trained using the training dataset. The deep learning model is provided for subsequent super-resolution image processing.

Different from prior works that model the internal distribution of patches within an image implicitly with a top-down latent variable model (e.g., generator), embodiments explicitly represent the statistical distribution within a single image by using an energy-based generative framework, where a pyramid of energy functions, each parameterized by a bottom-up deep neural network, are used to capture the distributions of patches at different resolutions. Also, embodiments of a coarse-to-fine sequential training and sampling strategy are presented to train the model efficiently. Besides learning to generate random samples from white noise, embodiments can learn in parallel with a self-supervised task (e.g., recover an input image from its corrupted version), which can further improve the descriptive power of the learned model. Embodiments does not require an auxiliary model (e.g., discriminator) to assist the training, and embodiments also unify internal statistics learning and image generation in a single framework.

This disclosure describes methods, non-transitory computer readable storage media, and systems that generate synthetized digital images via multi-resolution generator neural networks. The disclosed system extracts multi-resolution features from a scene representation to condition a spatial feature tensor and a latent code to modulate an output of a generator neural network. For example, the disclosed systems utilizes a base encoder of the generator neural network to generate a feature set from a semantic label map of a scene. The disclosed system then utilizes a bottom-up encoder to extract multi-resolution features and generate a latent code from the feature set. Furthermore, the disclosed system determines a spatial feature tensor by utilizing a top-down encoder to up-sample and aggregate the multi-resolution features. The disclosed system then utilizes a decoder to generate a synthesized digital image based on the spatial feature tensor and the latent code.

A method (and system) of segmenting one or more histological structures in a tissue image represented by multi-parameter cellular and sub-cellular imaging data includes receiving coarsest level image data for the tissue image, wherein the coarsest level image data corresponds to a coarsest level of a multiscale representation of first data corresponding to the multi-parameter cellular and sub-cellular imaging data. The method further includes breaking the coarsest level image data into a plurality of non-overlapping superpixels, assigning each superpixel a probability of belonging to the one or more histological structures using a number of pre-trained machine learning algorithms to create a probability map, extracting an estimate of a boundary for the: one or more histological structures by applying a contour algorithm to the probability map, and using the estimate of the boundary to generate a refined boundary for the one or more histological structures.

Disclosed are multiple object detection method and apparatus. The multiple object detection apparatus includes a feature map extraction unit for extracting a plurality of multi-scale feature maps based on an input image, and a feature map fusion unit for generating a multi-scale fusion feature map including context information by fusing adjacent multi-scale feature maps among the plurality of multi-scale feature maps generated by the feature map extraction unit.

A method and apparatus for establishing an image recognition model, a device, and a storage medium are provided. The method includes: acquiring an inputted image set; performing co-training on an initial super-resolution model and an initial recognition model using the inputted image set, to obtain a trained super-resolution model and a trained recognition model; and combining the trained super-resolution model and the trained recognition model in a cascaded manner to obtain the image recognition model.

A method of processing an image of a scene including a road acquired by a vehicle-mounted camera to generate boundary data indicative of a boundary of an image region which represents an unoccupied area of the road, comprising: generating (S10) an LL sub-band image of an N.sup.th level of an (N+1)-level discrete wavelet transform, DWT, decomposition of the image by iteratively low-pass filtering and down-sampling the image N times, where N is an integer equal to or greater than one; generating (S20) a sub-band image of an (N+1).sup.th level by high-pass filtering the LL sub-band image of the N.sup.th level, and down-sampling a result of the high-pass filtering, such that the sub-band image of the (N+1).sup.th level has a pixel region having substantially equal pixel values representing the unoccupied area of the road in the image; and generating (S30) the boundary data by determining a boundary of the pixel region.

A method of processing an image of a scene including a road acquired by a vehicle-mounted camera to generate boundary data indicative of a boundary of an image region which represents an unoccupied area of the road, comprising: generating (S10) an LL sub-band image of an N.sup.th level of an (N+1)-level discrete wavelet transform, DWT, decomposition of the image by iteratively low-pass filtering and down-sampling the image N times, where N is an integer equal to or greater than one; generating (S20) a sub-band image of an (N+1).sup.th level by high-pass filtering the LL sub-band image of the N.sup.th level, and down-sampling a result of the high-pass filtering, such that the sub-band image of the (N+1).sup.th level has a pixel region having substantially equal pixel values representing the unoccupied area of the road in the image; and generating (S30) the boundary data by determining a boundary of the pixel region.