International Business Machines Corporation

Xilinx, Inc.

Western Digital Technologies, Inc.

Globalfoundries Inc.

Hgst Netherlands, B.v.

David Berman

Wayne Isami Imaino

Daniel James Winarski

Dylan J Boday

Icko E T Iben

Anna Wanda Topol

Ephrem C Wu

Stephen Leonard Schwartz

William Marvin Dyer

Xiaoqian Zhang

Toshiki Hirano

Constantin Michael Melas

Glen Alan Jaquette

Robert Allen Hutchins

A David Erpelding

Alshakim Nelson

Pierre-Olivier Jubert

Shaomin Xiong

W Stanley Czarnecki

Ronald Robert Labby

Qiu Dai

Robert G Biskeborn

Nhan Xuan Bui

Sedat Oelcer

Chandrasekhar Narayan

Mario Blaum

Michel Despont

Philipp Herget

Satoshi Yamamoto

Rajeev Nagabhirava

Peter VanderSalm Koeppe

Mark Allan Taylor

Abhishek Dhanda

Eric Rolf Christensen

Hareem Tariq Maune

Haoyu Wu

Krishna M Bajjuri

Damien Kah

Qian Zhong

Icko Et Iben

Icko F Iben

Evangelos S Eleftherion

Michael Constantin Melas

Systems and methods for processing surveillance video streams using image classification and object detection are described. Video data from a video image sensor may be processed using an image classifier to determine whether an object type is present in a video frame. If the object type is present, the video frame and/or subsequent video frames may be processed using an object detector to provide additional object data, such as position information, for use in other video surveillance processes. In some examples, an event message may be generated and sent to a video surveillance application in response to selective object detection.

Processing of surveillance video streams using image classification and object detection

A device configured for more efficiently processing video images within a set of video image data to detect objects is described herein. The device may include a processor configured to execute a neural network such as a convolutional neural network. The device can receive video image data from a plurality of cameras, such as stationary cameras. The device can acquire a set of sample images from a stationary camera and submit them to a specialized neural network for processing to generate an attention mask. The attention mask can be generated from a variety of methods and is applied to each of the subsequently acquired images form the camera to narrow down areas where the convolutional neural network should process data. The application of attention masks to images within video image data creates masked images that can be processed to detect objects with much greater accuracy and fewer computational resources required.

Attention masks in neural network video processing

Embodiments herein describe circuitry with improved efficiency when executing layers in a nested neural network. As mentioned above, a nested neural network has at least one split operation where a tensor generated by a first layer is transmitted to, and processed by several branches in the neural network. Each of these branches can have several layers that have data dependencies which result in a multiply-add array sitting idly. In one embodiment, the circuitry can include a dedicated pre-pooler for performing a pre-pooling operation. Thus, the pre-pooling operation can be performing in parallel with other operations (e.g., the convolution performed by another layer). Once the multiply-add array is idle, the pre-pooling operation has already completed (or at least, has already started) which means the time the multiply-add array must wait before it can perform the next operation is reduced or eliminated.

Neural-network pooling

Systems and methods for surveillance video streams with embedded object data from another video camera are described. At least two video cameras are configured with fields of view to provide images of an object from alternative views. Video data for a primary video stream is received from one camera and secondary object data for the object from the other camera is embedded in the primary video stream. The primary video stream is sent to an analytics engine for processing the primary video and embedded secondary object data, such as performing facial recognition on a better image of a human face and/or feature vectors therefrom that are embedded in the primary video stream.

Surveillance video streams with embedded object data

Disclosed circuits and methods involve a first register configured to store of a first convolutional neural network (CNN) instruction during processing of the first CNN instruction and a second register configured to store a second CNN instruction during processing of the second CNN instruction. Each of a plurality of address generation circuits is configured to generate one or more addresses in response to an input CNN instruction. Control circuitry is configured to select one of the first CNN instruction or the second CNN instruction as input to the address generation circuits.

Neural network controller

A circuit arrangement includes an array of MAC circuits, wherein each MAC circuit includes a cache configured for storage of a plurality of kernels. The MAC circuits are configured to receive a first set of data elements of an IFM at a first rate. The MAC circuits are configured to perform first MAC operations on the first set of the data elements and a first one of the kernels associated with a first OFM depth index during a first MAC cycle, wherein a rate of MAC cycles is faster than the first rate. The MAC circuits are configured to perform second MAC operations on the first set of the data elements and a second one of the kernels associated with a second OFM depth index during a second MAC cycle that consecutively follows the first MAC cycle.

Performing consecutive mac operations on a set of data using different kernels in a MAC circuit

An integrated circuit (IC) includes a plurality of dies. The IC includes a plurality of memory channel interfaces configured to communicate with a memory, wherein the plurality of memory channel interfaces are disposed within a first die of the plurality of dies. The IC may include a compute array distributed across the plurality of dies and a plurality of remote buffers distributed across the plurality of dies. The plurality of remote buffers are coupled to the plurality of memory channels and to the compute array. The IC further includes a controller configured to determine that each of the plurality of remote buffers has data stored therein and, in response, broadcast a read enable signal to each of the plurality of remote buffers initiating data transfers from the plurality of remote buffers to the compute array across the plurality of dies.

Data transfers between a memory and a distributed compute array

A system, according to one embodiment, includes: a plurality of tape reels; and a tape drive configured for reading data from tape stored on at least one of the plurality of tape reels. At least some of the tape reels have a tape wound thereon and a spring-like clip coupled to a free end of the tape. Moreover, the clip is selectively positionable in a wrapped position where the clip wraps around a portion of the tape when the tape is wound onto the reel, thereby holding the portion of the tape in place on the reel.

Spring clip leader and housing for magnetic tape

The present invention relates to a water-soluble polymer complex that includes a water-soluble block copolymer and a magnetic nanoparticle, wherein the water-soluble polymer complex has a nonzero net magnetic moment in the absence of an applied magnetic field at ambient temperatures. The water-soluble block copolymer is preferably a diblock or triblock copolymer and the magnetic nanoparticle is preferably a ferrimagnetic or ferromagnetic nanoparticle. The water-soluble complexes may be derivatized with reactive groups and conjugated to biomolecules. Exemplary water-soluble polymer complexes covered under the scope of the invention include PEG-b-PAAmodified CoFeO; NH-PEG-b-PAAmodified CoFeO; PNIPAM-b-PAAmodified CoFeO; and mPEG-b-PCL-b-PAA modified CoFeO.

Aqueous soluble ferrimagnets stabilized by block copolymers

Described are embodiments to calibrate read sensors, which in turn may ensure that the equipment utilized to detect antigens is reliable and accurate. If it is determined that a read sensor is degraded a method of calibrating a read sensor of a read head may be used.

Calibrating read sensors of electromagnetic read-write heads

Disclosed circuitry includes RAM circuits, a memory controller, and an array of processing circuits. Each RAM circuit includes a read port and a write port. The memory controller accesses tensor data arranged in banks of tensor buffers in the RAM circuits. The memory controller is coupled to each read port by shared read control signal lines and to each write port by shared write control signal lines. The memory controller generates read control and write control signals for accessing different ones of the tensor buffers at different times. The array of processing circuits is coupled to one of the RAM circuits. The array includes multiple rows and multiple of columns of processing circuits for performing tensor operations on the tensor data. The processing circuits in each row in each array of processing circuits are coupled to input the same tensor data.

Memory arrangement for tensor data

The examples described herein discuss various systems, software, devices, and methods for managing a primary command queue by ordering and/or reordering and distributing incoming commands based, at least in part, on positional information of one or more components of data storage devices. More specifically, in some embodiments, the examples discussed herein describe ordering and distributing incoming commands from a primary command queue in a position-aware manner that takes into account disk rotation (e.g., rotational position) and/or actuator head location for the various data storage devices of a data storage system enclosure. Among other benefits, ordering incoming commands at the primary command queue and distributing the ordered commands to individual device queues improves overall command execution latency.

Position-aware primary command queue management

A spring-like clip for a spool of tape, one end of the clip being attached to one end of the tape, and another end of the clip having an engagement feature that allows the clip to be pulled from the spool. The clip is configured to curve around the spool to hold the tape in place. In one general embodiment, a product includes a reel, a tape coupled to the reel, and a spring-like clip coupled to a free end of the tape. The clip is selectively positionable in a wrapped position where the clip wraps around a portion of the tape when the tape is wound onto the reel, thereby holding the portion of the tape in place on the reel. Systems for using various types of such products are presented, and generally include a tape drive configured for reading data from tape stored on such products.

According to one embodiment, a calibration assembly includes an outer layer having at least one calibration trench extending along a y-axis, and an encapsulation layer within the calibration trench. The encapsulation layer has a plurality of nanoparticles spaced apart along said y-axis of said at least one calibration trench. Each of said plurality of nanoparticles are provided at known y-axis locations in said calibration trench, and each of the plurality of nanoparticles have a known magnetic property.

Calibration assembly for aide in detection of analytes with electromagnetic read-write heads

According to one embodiment a method of performing a calibration correlation test for a calibration assembly includes sweeping a head module having a magnetic read sensor along a y-axis of the calibration assembly. The calibration assembly has at least one calibration trench having at least one nanoparticle at a known y-axis location in the calibration trench and the magnetic properties are known for the at least one nanoparticle. A read response of the at least one nanoparticles is obtained from the magnetic read sensor and a correlation is determined from the read response. The correlation of the read response is compared to a correlation threshold. The read response correlation is stored in memory in response to determining that the correlation of the read response is greater than the correlation threshold. When the correlation of the read response is not greater than the correlation threshold, a correlation test error is indicated.

Calibration correlation for calibration assembly having electromagnetic read head

Growing community of inventors

San Jose, CA, United States of America

David Berman