Nokia Technologies Oy

Inception Institute of Artificial Intelligence, Ltd.

Jiale Cao

Ling Shao

Fahad Shahbaz Khan

Hisham Cholakkal

Rao Muhammad Anwer

Yanwei Pang

A method for pattern recognition may be provided, comprising: receiving data; processing the data with a convolutional neural network in order to recognize a pattern in the data, wherein the convolutional neural network comprises at least: a first branch comprising a sequence of first convolutional blocks, a pooling layer being disposed between any two adjacent first convolutional blocks, each first convolutional blocks comprising at least one convolutional layer, and a second branch comprising a sequence of second convolutional blocks, each second convolutional blocks comprising at least one convolutional layer, and wherein processing the data with a convolutional neural network in order to recognize a pattern in the data comprises: a preceding second convolutional block receiving a first feature map formed by combining of a feature map outputted by a preceding first convolutional block and a feature map outputted by a subsequent first convolutional block, processing the first feature map, and outputting a second feature map, and a subsequent second convolutional block receiving a third feature map formed by combining the second feature map and a feature map outputted by a further subsequent first convolutional block, processing the third feature map, and outputting a fourth feature map. A corresponding apparatus and system for pattern recognition, as well as a computer readable medium, are also provided.

Method, system and apparatus for pattern recognition

Various methods are provided for training and subsequently utilizing a convolutional neural network (CNN) to detect small pedestrians (e.g., pedestrians located away a large distance). One example method may comprise performing a first training stage in which a first CNN is trained to detect objects of a first size, the first CNN trained using a first set of images comprised of objects of the first size, and configured to output a first set of parameters, performing a second training stage in which a second CNN is trained using a second set of images, the second set of images comprising objects of a second size, and the first CNN is initialized with the first set of parameters and is re-trained using the second set of images, and determining parameters of the first CNN by minimizing error between the first CNN and the second CNN.

Method and apparatus for detecting small objects with an enhanced deep neural network

A method for pattern recognition may be provided, comprising: receiving data; processing the data with a trained convolutional neural network so as to recognize a pattern in the data, wherein the convolutional neural network comprises at least: an input layer, at least one convolutional layer, at least one batch normalization layer, at least one activation function layer, and an output layer; and wherein processing the data with a trained convolutional neural network so as to recognize a pattern in the data comprises: processing values outputted by a batch normalization layer so that the histogram of the processed values is flatter than the histogram of the values, and outputting the processed values to an activation function layer. A corresponding apparatus and system for pattern recognition, as well as a computer readable medium, a method for implementing a convolutional neural network and a convolutional neural network are also provided.

This disclosure relates to improved techniques for performing computer vision functions, including common object detection and instance segmentation. The techniques described herein utilize neural network architectures to perform these functions in various types of images, such as natural images, UAV images, satellite images, and other images. The neural network architecture can include a dense location regression network that performs object localization and segmentation functions, at least in part, by generating offset information for multiple sub-regions of candidate object proposals, and utilizing this dense offset information to derive final predictions for locations of target objects. The neural network architecture also can include a discriminative region-of-interest (RoI) pooling network that performs classification of the localized objects, at least in part, by sampling various sub-regions of candidate proposals and performing adaptive weighting to obtain discriminative features.

Dense and discriminative neural network architectures for improved object detection and instance segmentation

An apparatus comprises memory configured to store, at least partly, labelling information of a convolutional artificial neural network, and at least one processing core configured to generate, from an input data item, partial feature maps of the convolutional artificial neural network in accordance with the labelling information, generate, from the partial feature maps, inputs to a plurality of weak classifiers to generate a classification decision, wherein the labelling information identifies at least one of the following: elements of the feature maps that generate the inputs, and elements of the feature maps that are used to generate the elements that generate the inputs.

Artificial neural network

Methods, systems and apparatuses of feature extraction and object detection are provided. In the method of feature extraction, a plurality of image channels are generated from each of training images; intra-channel features are extracted from the plurality of image channels for each of training images, wherein the intra-channel features include features independently extracted from a single image channel; cross-channel features are extracted from the plurality of image channels for at least one of the training images, wherein the cross-channel features include features extracted from at least two image channels. The intra-channel features and the cross-channel features form a set of features for feature selection and classifier training. With the above method, cross-channel features, which reflect discriminant information across different image channels, can be further used for object detection together with the intra-channel features, and thus there are much richer features for object detection and better accuracy of object detection can be achieved.

Growing community of inventors

Tianjin, China

Jiale Cao