Abstract: Disclosed is a processor-implemented data processing method in a neural network. A data processing apparatus includes at least one processor, and at least one memory configured to store instructions to be executed by the processor and a neural network, wherein the processor is configured to, based on the instructions, input an input activation map into a current layer included in the neural network, output an output activation map by performing a convolution operation between the input activation map and a weight quantized with a first representation bit number of the current layer, and output a quantized activation map by quantizing the output activation map with a second representation bit number based on an activation quantization parameter.

Abstract: A method and apparatus with object tracking is provided. The method includes generating a mixed filter by fusing a short-term filter with a long-term filter; and performing object tracking on a current frame image based on the mixed filter. The short-term filter is dependent on a prediction of the current frame image in a video sequence, and the long-term filter is a previously generated long-term filter or is generated by optimizing the previously generated long-term filter based on an object template feature pool.

Abstract: A processor-implemented method of generating feature data includes: receiving an input image; generating, based on a pixel value of the input image, at least one low-bit image having a number of bits per pixel lower than a number of bits per pixel of the input image; and generating, using at least one neural network, feature data corresponding to the input image from the at least one low-bit image.

Abstract: A processor-implemented method with object tracking includes: performing, using a first template, forward object tracking on first image frames in a first sequence group; determining a template candidate of a second template for second image frames in a second sequence group; performing backward object tracking on the first image frames using the template candidate; determining a confidence of the template candidate using a result of comparing a first tracking result determined by the forward object tracking performed on the first image frames and a second tracking result determined by the backward object tracking performed on the first image frames; determining the second template based on the confidence of the template candidate; and performing forward object tracking on the second image frames using the second template.

Abstract: A processor-implemented method of tracking a target object includes: extracting a feature from frames of an input image; selecting one a neural network model from among a plurality of neural network models that is provided in advance based on a feature value range, based on a feature value of a target object that is included in the feature of a previous frame among the frames; and generating a bounding box of the target object included in a current frame among the frames, based on the selected neural network model.

Abstract: Disclosed is a method and apparatus for face detection using an adaptive threshold. The method includes determining a detection box in an input image, calculating a confidence score indicating whether an object in the detection box corresponds to a face, setting an adaptive threshold based on a size of the detection box, and determining whether the object in the detection box corresponds to a face based on comparing the confidence score to the adaptive threshold.

Abstract: A processor-implemented method of generating feature data includes: receiving an input image; generating, based on a pixel value of the input image, at least one low-bit image having a number of bits per pixel lower than a number of bits per pixel of the input image; and generating, using at least one neural network, feature data corresponding to the input image from the at least one low-bit image.

Abstract: A method and apparatus for object tracking are provided, where the object tracking method includes determining box information of candidate boxes in a current image frame and similarity scores of the candidate boxes based on including a search region of the current image frame with a template image corresponding to a target object, adjusting the similarity scores of the candidate boxes using a distractor map including distractor information of a previous image frame, determining a target box corresponding to the target object and a distractor box corresponding to a distractor of the target object from the candidate boxes based on the adjusted similarity scores, and updating the distractor map based on distractor information of the current image frame according to the distractor box.

Abstract: Disclosed is a method and apparatus for face detection using an adaptive threshold. The method includes determining a detection box in an input image, calculating a confidence score indicating whether an object in the detection box corresponds to a face, setting an adaptive threshold based on a size of the detection box, and determining whether the object in the detection box corresponds to a face based on comparing the confidence score to the adaptive threshold.

Abstract: A method with self-attention includes: obtaining a three-dimensional (3D) feature map; generating 3D query data and 3D key data by performing a convolution operation based on the 3D feature map; generating two-dimensional (2D) vertical data based on a vertical projection of the 3D query data and the 3D key data; generating 2D horizontal data based on a horizontal projection of the 3D query data and the 3D key data; determining an intermediate attention result through a multiplication based on the 2D vertical data and the 2D horizontal data; and determining a final attention result through a multiplication based on the intermediate attention result and the 3D feature map.

Abstract: A processor-implemented method of generating feature data includes: receiving an input image; generating, based on a pixel value of the input image, at least one low-bit image having a number of bits per pixel lower than a number of bits per pixel of the input image; and generating, using at least one neural network, feature data corresponding to the input image from the at least one low-bit image.

Abstract: A processor-implemented neural network processing method includes: obtaining a kernel bit-serial block corresponding to first data of a weight kernel of a layer in a neural network; generating a feature map bit-serial block based on second data of one or more input feature maps of the layer; and generating at least a portion of an output feature map by performing a convolution operation of the layer using a bitwise operation between the kernel bit-serial block and the feature map bit-serial block.

Abstract: A processor-implemented method of tracking a target object includes: extracting a feature from frames of an input image; selecting one a neural network model from among a plurality of neural network models that is provided in advance based on a feature value range, based on a feature value of a target object that is included in the feature of a previous frame among the frames; and generating a bounding box of the target object included in a current frame among the frames, based on the selected neural network model.

Abstract: An object tracking method includes generating a feature map of a search image and generating a feature map of a target image, obtaining an object classification result and a basic bounding box based on the feature map of the search image and the feature map of the target image, obtaining an auxiliary bounding box based on the feature map of the search image, obtaining a final bounding box based on the basic bounding box and the auxiliary bounding box, and tracking an object based on the object classification result and the final bounding box.

Abstract: Disclosed is a method and apparatus for adaptive tracking of a target object. The method includes method of tracking an object, the method including estimating a dynamic characteristic of an object in an input image based on frames of the input image, determining a size of a crop region for a current frame of the input image based on the dynamic characteristic of the object, generating a cropped image by cropping the current frame based on the size of the crop region, and generating a result of tracking the object for the current frame using the cropped image.

Abstract: An object tracking method includes generating a feature map of a search image and generating a feature map of a target image, obtaining an object classification result and a basic bounding box based on the feature map of the search image and the feature map of the target image, obtaining an auxiliary bounding box based on the feature map of the search image, obtaining a final bounding box based on the basic bounding box and the auxiliary bounding box, and tracking an object based on the object classification result and the final bounding box.

Abstract: A processor-implemented method of generating feature data includes: receiving an input image; generating, based on a pixel value of the input image, at least one low-bit image having a number of bits per pixel lower than a number of bits per pixel of the input image; and generating, using at least one neural network, feature data corresponding to the input image from the at least one low-bit image.

Abstract: A processor-implemented neural network data processing method includes: determining a total number of either one of a first feature value and values less than or equal to the first feature value, in feature data output from a layer of a neural network; determining a quantization parameter based on the determined number; quantizing the feature data based on the determined quantization parameter; and inputting the quantized feature data to a another layer of the neural network connected to the layer.

Abstract: A processor-implemented method of generating feature data includes: receiving an input image; generating, based on a pixel value of the input image, at least one low-bit image having a number of bits per pixel lower than a number of bits per pixel of the input image; and generating, using at least one neural network, feature data corresponding to the input image from the at least one low-bit image.

Abstract: Disclosed are methods and apparatuses for processing a convolution operation on a layer in a neural network. The method includes extracting a first target feature vector from a target feature map, extracting a first weight vector matched with the first target feature vector from a first-type weight element, based on matching relationships for depth-wise convolution operations between target feature vectors of the target feature map and weight vectors of the first-type weight element, generating a first intermediate feature vector by performing multiplication between the first target feature vector and the first weight vector, generating a first hidden feature vector by accumulating the first intermediate feature vector and a second intermediate feature vector generated based on a second target feature vector, and generating a first output feature vector of an output feature map based on a point-wise convolution operation between the first hidden feature vector and a second-type weight element.