An eye location method and device includes: receiving a face image; locating a position of a nose and positions of eyes in the face image; determining a facial symmetry axis according to the position of the nose; and regulating the positions of the eyes by the facial symmetry axis to obtain target positions of the eyes. By the method and the device, the problem of poorer eye location accuracy in the conventional art is solved, and the effect of improving eye location accuracy is further achieved.

Systems, apparatuses, and methods for converting data to a tiling format when implementing convolutional neural networks are disclosed. A system includes at least a memory, a cache, a processor, and a plurality of compute units. The memory stores a first buffer and a second buffer in a linear format, where the first buffer stores convolutional filter data and the second buffer stores image data. The processor converts the first and second buffers from the linear format to third and fourth buffers, respectively, in a tiling format. The plurality of compute units load the tiling-formatted data from the third and fourth buffers in memory to the cache and then perform a convolutional filter operation on the tiling-formatted data. The system generates a classification of a first dataset based on a result of the convolutional filter operation.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating numeric embeddings of images. One of the methods includes obtaining training images; generating a plurality of triplets of training images; and training a neural network on each of the triplets to determine trained values of a plurality of parameters of the neural network, wherein training the neural network comprises, for each of the triplets: processing the anchor image in the triplet using the neural network to generate a numeric embedding of the anchor image; processing the positive image in the triplet using the neural network to generate a numeric embedding of the positive image; processing the negative image in the triplet using the neural network to generate a numeric embedding of the negative image; computing a triplet loss; and adjusting the current values of the parameters of the neural network using the triplet loss.

Techniques describe computing computer vision (CV) features based on sensor readings from a sensor and detecting macro-features based on the CV features. The sensor may include a sensor element array that includes a plurality of sensor elements. The sensor may also include in-pixel circuitry coupled to the sensor elements, peripheral circuitry and/or a dedicated microprocessor coupled to the sensor element array. The in-pixel circuitry, the peripheral circuitry or the dedicated microprocessor may include computation structures configured to perform analog or digital operations representative of a multi-pixel computation for a sensor element (or block of sensor elements), based on sensor readings generated by neighboring sensor elements in proximity to the sensor element, and to generate CV features. The dedicated microprocessor may process the CV features and detect macro-features. Furthermore, in certain embodiments, the dedicated microprocessor may be coupled to a second microprocessor through a wired or wireless interface.

This application provides a skin color detection method and apparatus. The skin color detection method includes: obtaining a face image (101); determining a face key point (102) in the face image; determining a skin color estimation region of interest (Region Of Interest, ROI) and an illumination estimation region of interest ROI (103) in the face image based on the face key point; obtaining a detected skin color value (104) corresponding to the skin color estimation region of interest; obtaining a detected illumination color value (105) corresponding to the illumination estimation region of interest; and using the detected skin color value and the detected illumination color value as feature input of a skin color estimation model, and obtaining a corrected skin color value (106) output by the skin color estimation model.

Embodiments relate to generating an image pyramid for feature extraction. A pyramid image generator circuit includes a first image buffer that stores image data at a first octave, a first blur filter circuit, a first spatial filter circuit, and a first decimator circuit. The first blur filter circuit generates a first pyramid image for a first scale of the first octave by applying a first amount of smoothing to the first image data stored in the first image buffer. The first spatial filter circuit and the first decimator generate second image data of a second octave that is higher than the first octave by applying a smoothing and a decimation to the first image data stored in the first image buffer. The first spatial filter circuit begins processing the first image data before the first blur filter circuit begins to process the first image data.

The present invention discloses a method for finding a cell nucleus of a target cell from a cell image, wherein the cell image includes the target cell and at least one variation cell, and the target cell includes cytoplasm and the cell nucleus. The method includes steps of: (a) processing the cell image via an image processor such that the cytoplasm, the cell nucleus and the variation cell have different shades of color; (b) demarcating the outlines of the cytoplasm, the cell nucleus and the variation cell; (c) calculating geometrical reference points of the outlines; (d) calculating the distances from the geometrical reference point of the cytoplasm outline to the geometrical reference point of the cell nucleus outline and to the geometrical reference points of the variation cell outlines; and (e) finding a specific geometrical reference point having a shortest distance to locate a specific outline corresponding to the specific geometrical reference point as the cell nucleus.

Provided us a visual cortical circuit apparatus comprising: a current mirror unit which uses a transistor as a current source to generate a current having the same size as that of a reaction; a transconductance unit which takes, as an input, the current generated by the current mirror unit and outputs a voltage using a transconductance; and a buffer unit for converting the voltage output from the transconductance unit into a current and buffering the current.

A machine learning (ML) task system trains a neural network model that learns a compressed representation of acquired data and performs a ML task using the compressed representation. The neural network model is trained to generate a compressed representation that balances the objectives of achieving a target codelength and achieving a high accuracy of the output of the performed ML task. During deployment, an encoder portion and a task portion of the neural network model are separately deployed. A first system acquires data, applies the encoder portion to generate a compressed representation, performs an encoding process to generate compressed codes, and transmits the compressed codes. A second system regenerates the compressed representation from the compressed codes and applies the task model to determine the output of a ML task.

A device to extract a biometric feature vector includes a memory and a circuitry. The circuitry is configured to obtain a biometric image, to generate a plurality of small region images from the biometric image so that variability of biometric information amounts among the plurality of small region images is equal to or less than a predetermined value, to extract biometric local feature amounts from the small region images and to generate a biometric feature vector by combining the biometric local feature amounts in accordance with a predetermined rule, the biometric feature vector indicating a feature for identifying the biometric image.