Fujitsu Corporation

Zhiming Tan

Jing Zhou

Sihan Wen

Rui Yin

Lan Ding

Akira Nakagawa

Xianghui Bai

Xiantan Zhu

Yasushi Sugama

Jianrong Wu

Akihiro Higashi

Xiaoqing Liu

Bingrong Wang

Embodiments of this disclosure provide an apparatus and method. As target detection is performed in the detected pedestrian area by directly using the deep learning model, various article belonging to a pedestrian can be accurately detected in an input image.

Pedestrian article detection apparatus and method and electronic device

Embodiments of this disclosure provide an image encoding method and apparatus and image decoding method and apparatus. The image encoding includes performing convolutional neural network (CNN) encoding on image data to generate feature vectors or feature maps; quantizing the feature vectors or feature maps to generate discrete symbols to be encoded; and estimating probabilities of the symbols to be encoded by using a multi-scale context model including multiple mask convolution layers of different scales. An entropy encoding of the image data is performed according to the probabilities of the symbols to be encoded.

Image encoding method and apparatus and image decoding method and apparatus

Embodiments of this disclosure provide an action recognition method and apparatus and an electronic equipment. The action recognition apparatus includes: a key-point feature extracting unit configured to process information on key points in a target human body in an image frame, to calculate key-point features of the target human body, the key-point features including a position of a predetermined key point and an angle of a predetermined joint in the target human body; and a first recognizing unit configured to recognize an action of the target human body according to the key-point features, and output a first recognition result.

Action recognition method and apparatus and electronic equipment

Embodiments of this disclosure provide an image coding apparatus, a probability model generating apparatus and an image decoding apparatus. A processor is to perform feature extraction on an input image to obtain first feature maps of N channels; to perform feature extraction on the input image with a size of the input image being adjusted K times, to respectively obtain second feature maps of N channels; and to concatenate the first feature maps of the K&#xd7;N channels with the second feature maps of K&#xd7;N channels to output a concatenated feature maps of channels. Hence, features of images may be accurately extracted and more competitive latent representations may be obtained.

Image coding apparatus, probability model generating apparatus and image decoding apparatus

Embodiments of this disclosure provide a video frame interpolation apparatus and method. The method includes calculating a bidirectional optical flow between a first frame and a second frame; an performing kernel and weight estimation according to the first frame and the second frame. An adaptive local convolutional kernel is generated by using a convolutional layer and a weight coefficient is generated by using another convolutional layer. A conversion on the first frame and the second frame is performed by using an adaptive conversion layer according to the bidirectional optical flow, the weight coefficient and the adaptive local convolutional kernel, so as to generate a conversion frame. A frame synthesis on the first frame, the second frame and the conversion frame is performed to generate an interpolation frame between the first frame and the second frame.

Video frame interpolation apparatus and method

Embodiments of this disclosure provide a deep learning model used for image recognition and apparatus and method thereof. The model includes a determination layer configured to determine whether features in feature maps are features of positions where objects of attention are located, and different weights are granted for the positions where the objects of attention are located and other features in performing weight and composition processing on the features. Hence, the model may be guided to be focused on attention features and make correct determination, thereby improving performance and precision of the model.

Deep learning model used for image recognition and training apparatus of the model and method thereof

Embodiments of this disclosure provide a training method, an image encoding method, an image decoding method and apparatuses thereof. The image encoding apparatus includes: an image encoder configured to encode input image data to obtain a latent variable; a quantizer configured to perform quantizing processing on the latent variable according to a quantization step to generate a quantized latent variable; and an entropy encoder configured to perform entropy coding on the quantized latent variable by using an entropy model to form a bit stream.

Training method, image encoding method, image decoding method and apparatuses thereof

Embodiments of this disclosure provide an arm action identification method and apparatus and an image processing device. A human body detection is performed on an image to obtain multiple key points of the human body in the image. An arm action is identified to be an upward action or a downward action by using a shoulder key point, an elbow key point and a wrist key point in the multiple key points. Arm actions indicative of upward, downward, or forward, actions may be identified based upon, for example, lifting, putting down, holding forward, actions, by using the key points of the human body, thereby achieving computer-level action identification, and lowering complexity of computation.

Arm action identification method and apparatus and image processing device

Embodiments of this disclosure provide an image coding method and apparatus and an image compression system. The image coding apparatus includes a memory and a processor. The processor is configured to perform feature extraction on an input image to obtain feature maps of N channels; assign a weight to a feature map of each channel among the N channels; perform down-dimension processing on weighted feature maps processed in association with the N channels, to obtain feature maps of M channels and output the feature maps of M channels, M being smaller than N. Hence, by multiplying different feature maps by a weight to obtain corresponding importance and then performing down-dimension processing on the feature maps processed according to the weighting, time for decoding may be reduced.

Image coding apparatus, probability model generating apparatus and image compression system

Embodiments of this disclosure provide a deep learning model used for driving behavior recognition and training apparatus and method thereof. In recognition, the model performs feature extraction by using a plurality of consecutive input images captured from a driving angle of view of a vehicle, and performs temporal and spatial fusion on extracted features by using a recursive neural network. In training, as images captured from a driving angle of view of a vehicle are acquired, the model may be trained. Hence, the model may accurately recognize various classes of driving behaviors.

Deep learning model used for driving behavior recognition and training apparatus and method thereof

Embodiments of this disclosure provide a smoke detection method and apparatus. The apparatus includes a processor configured to detect foreground blobs in images of a plurality of frames in a video; determine motion areas of detected foreground blobs; determine a first degree of overlapping of the foreground blobs in images of at least two frames, and/or determine a second degree of overlapping of the foreground blobs and motion areas to which the foreground blobs correspond; determine interfering foreground blobs according to the first degree of overlapping and/or the second degree of overlapping; and extract features from motion areas of remaining foreground blobs with the interfering foreground blobs being removed, and detect smoke according to the features. With the embodiments of this disclosure, accuracy of smoke detection may be increased, and false detection resulted from such interfering objects as water blobs, or light spots, etc., may be avoided.

Smoke detection method and apparatus

Embodiments of this disclosure provide an image coding method and apparatus and an image decoding method and apparatus. The image coding method includes: performing feature extraction on to-be-processed image data by using a convolutional neural network, to generate feature maps of the image data; quantizing the feature maps to generate discrete feature maps; preprocessing the discrete feature maps to generate preprocessed data, an amount of data of the preprocessed data being less than an amount of data of the discrete feature maps; calculating probabilities of to-be-coded data in the discrete feature maps according to the preprocessed data; and performing entropy coding on the to-be-coded data according to the probabilities of the to-be-coded data.

Image coding method and apparatus and image decoding method and apparatus

Embodiments of this disclosure provide a person searching method and apparatus and an image processing device. The method includes: performing person detection on a reference image, determining a query person, and acquiring property information on the query person; performing person detection on at least one video surveillance image, determining all candidate persons, and acquiring property information on all the candidate persons; calculating a match score of each candidate person according to the property information on the query person and property information on each candidate person, and selecting a predetermined number of candidate persons in a descending order of the match scores; and calculating similarities between the query person and the selected candidate persons, and taking a candidate person in the selected candidate persons having a highest similarity with the query person as the query person. In the embodiments of this disclosure, the query person is found from the video surveillance image by person detection and two-step retrieval.

Person searching method and apparatus and image processing device

A method and a device for augmenting data in a monitoring video are provided which includes: selecting randomly, based on a function to be implemented for the data and a parameter range which are set by a user or by default, one or more processing functions to be performed on the data and parameters of the processing functions; processing the data using the selected one or more processing functions, to obtain new data; and repeating the above processes until a required amount of data is generated. With the method, monitoring data having diversity and a small size can be obtained with only a few input samples, and a high degree of randomness and comprehensiveness of the augmented data can be achieved.

Method and apparatus for augmenting data in monitoring video

An apparatus for image contrast enhancement and a method, electronic equipment that includes: a dividing unit configured to divide an input image into at least two layers; and detail degrees of the at least two layers are different; an enhancing unit configured to perform contrast enhancement on each of the at least two layers respectively; and degrees of enhancing contrasts of the layers are different; and a combining unit configured to combine the contrast-enhanced at least two layers. By dividing the input image into at least two layers with different detail degrees and performing contrast enhancement on the at least two layers at different degrees, the contrast of the image may be accurately and effectively enhanced, thereby a display effect of the image may be improved.

Growing community of inventors

Beijing, China

Zhiming Tan