International Business Machines Corporation

Sap Se

Chang Xu

Junsong Wang

Yan Gong

Tao Wang

Hang Liu

Chao Xue

Yubo Li

Yonghua Lin

Ke Wei Sun

Chao Zhu

Leonid Karlinsky

Qing Wang

Using a training portion of a dataset, a set of component parameters comprising parameters of a component of an object detection model are trained. Using the trained set of component parameters, a set of backbone component weights comprising weights of component types in a backbone portion of the object detection model are trained. Using the trained set of component parameters, a set of backbone link weights comprising weights of links within the backbone portion are trained. Using the trained set of component parameters, a set of head component weights comprising weights of component types in a head portion of the object detection model are trained. Using the trained sets of component parameters, backbone component weights, backbone link weights, and head component weights, a trained object detection model is configured and trained to perform object detection.

Automated end-to-end machine learning model optimization

Embodiments of the present disclosure include techniques for validating structured electronic documents. In one embodiment, a structured electronic document comprises a reference to a fragment in another structure electronic document. The documents are combined to replace reference with the fragment. The resulting combined document may be validated to generate a first diagnostic array corresponding to the combined document. The first diagnostic array is converted to a second diagnostic array corresponding to the original document. A validation result may specify where in the original documents an error occurred.

Systems and methods for validating structured documents

Methods, computer program products, and/or systems are provided that perform the following operations: obtaining a target neural network structure and constraints for a target neural network; generating a meta learning network having an associated quantization function based, at least in part, on the target neural network structure; training the meta learning network based, at least in part, on providing a hybrid quantization vector as input to the meta learning network and providing a training dataset to the target neural network; obtaining a plurality of hybrid quantization vectors; determining a new hybrid quantization vector from the plurality of hybrid quantization vectors; and retraining the trained meta learning network based, at least in part, on providing the new hybrid quantization vector as input to the trained meta learning network.

Automatic hybrid quantization for deep neural network

In some examples, a method includes receiving an artificial intelligence (AI) system scenario definition file from a user, parsing the definition file and building an application workflow graph for the AI system, and mapping the application workflow graph to an execution pipeline. In some examples, the method further includes automatically generating, from the workflow graph, application executable binary code implementing the AI system, and outputting the application executable binary code to the user. In some examples, the execution pipeline includes one or more building blocks, and the method then further includes collecting running performance of each of the building blocks of the execution pipeline in a runtime environment.

Artificial intelligence workflow builder

A method for configuring a set of hardware accelerators to process a CNN. In an embodiment, the method includes one or more computer processors determining a set of parameters related to a feature map to analyze at a respective layer of the CNN, the set of parameters include quantization value and respective values that describe a shape of the feature map. The method further includes configuring a set of hardware accelerators for the respective layer of the CNN. The method further includes receiving a portion of the feature map to the configured set of hardware accelerators for the respective layer of the CNN, wherein the received portion of the feature map includes a group of sequential data slices. The method further includes analyzing the group of sequential data slices among the configured set of hardware accelerators.

Accelerator resource utilization by neural networks

In an approach for detecting queuing information, a processor analyzes a video monitoring a queue area. A processor detects a queue barrier in the queue area using an instance segmentation technique based on the video. A processor identifies a queue in the queue area using a heuristic technique. A processor recognizes a number of people in the queue. A processor provides an estimation of a wait time for the queue.

Adaptive queue management system

A method, system, and computer program product for detecting multi-scale objects. The method may include receiving a sample dataset including multi-scale objects associated with a specific environment, where the sample dataset has an existing resolution. The method may also include inputting the sample dataset into a trained neural network, where the trained neural network has a plurality of scale regions. The method may also include processing the sample dataset for detecting the multi-scale objects by means of the trained neural network. The method may also include calculating a distribution of contribution degree in a course of processing the sample dataset, where the contribution degree is associated with each of the plurality of scale regions. The method may also include generating a set of configuration parameters associated with the specific environment for the trained neural network based at least in part on the distribution of contribution degree.

Multi-scale object detection with a trained neural network

A computer implemented method of detecting excessive customer wait times is provided. The method includes taking a headcount in a digital image obtained by a digital video camera of a monitored area, counting the number of bodies in the digital image, and rectifying the number of heads with the number of bodies to obtain a total count of persons. The method further includes determining which persons are moving and subtracting the moving persons from the total count of persons to obtain a still count, and determining which persons are workers and subtracting the workers from the still count to identify customers in the monitored area and obtain a customer count. The method further includes identifying the number of queues present in the monitored area, assigning each customer to a queue, and determining the wait time for each of the identified customers in each of the identified queues.

Growing community of inventors

Beijing, China

Chang Xu