International Business Machines Corporation

Korea Advanced Institute of Science and Technology

Chungkuk Yoo

Inseok Hwang

Eric Rozner

Su Liu

Myung-Chul Kim

Gelareh Taban

Yu Gu

Yaoguang Wei

Chin Ngai Sze

Janani Mukundan

Youngki Lee

Junehwa Song

Chulhong Min

Younghyun Ju

Uichin Lee

Seungwoo Kang

Miri Moon

Chanyou Hwang

Sangwon Choi

Jaeung Lee

Pillsoon Park

Seungpyo Choi

A method identifies a type of object in a digital image. A user and/or one or more processors selects, from a plurality of partially overlapping sub-images of a digital image, a first sub-image and a second sub-image that overlap one another. The user/processors input the first sub-image into a neural network to create a first inference result that includes an overlapping inference result, for the overlapping area, that recognizes a partial portion of a specific type of object based on the overlapping area. The user/processors infer that the second sub-image creates a second inference result that recognizes a second portion of the specific type of object in the second sub-image based on the second sub-image and the overlapping inference result. The neural network identifies the specific type of object in the digital image based on the first and second sub-images being sub-images of a same type of object.

Identifying a type of object in a digital image based on overlapping areas of sub-images

A method, computer program product, and a system where a processor(s), monitors multilingual switches performed on a client on behalf of a given user. Based on the monitoring, the processor(s) identifies switch patterns of the given user to generate a service profile for the user of machine learned multilingual switch patterns for the given user. The processor(s) determines a priority order for languages comprising the voice input streams, for the given user. The processor(s) obtains a new translation request initiated by the client, on behalf of the given user and applies the priority order to identify one or more languages spoken in a voice input stream of the new translation request. The processor(s) transmits indicators of the identified one or more languages to the client, where upon receiving the indicators, the client translates the voice input stream from the identified one or more languages to one or more target languages.

Dynamic multilingual speech recognition

A method optimizes Convolutional Neural Network (CNN) inference time for full resolution images. One or more processors divide a full resolution image into a plurality of partially overlapping sub-images. The processor(s) select, from the plurality of partially overlapping sub-images, a first sub-image and a second sub-image that overlap one another in an overlapping area. The processor(s) feed the first sub-image, including the overlapping area, into a Convolutional Neural Network (CNN) in order to create a first inference result for the first sub-image, where the CNN has been trained at a fine resolution. The processor(s) cache an inference result from the CNN for the overlapping area, and then utilize the cached inference result when inferring the second sub-image in the CNN. The processor(s) then identify a specific object in the full resolution image based on inferring the first sub-image and the second sub-image.

Overlapping CNN cache reuse in high resolution and streaming-based deep learning inference engines

A method includes: receiving light at an image sensor of a digital camera; generating an image based on the received light; detecting an artifact in the image; determining a blocking area of the sensor based on the detecting the artifact; blocking the light at the blocking area; determining whether the artifact is present in the image after the blocking; in response to determining that the artifact is present after the blocking, determining a new blocking area and repeating the blocking and the determining using the new blocking area; and in response to determining that the artifact is not present after the blocking, saving the image in memory.

Eliminating digital image artifacts caused by backlighting

Disclosed herein is a communication device, which comprises a user trace manager, a power emulator, and a power impact estimator. The user trace manager manages a user's sensor usage trace and device usage trace, which shows the user's mobile device usage pattern. The power emulator executes an executable file of the mobile application to generate hardware usage statistics based on the sensor usage trace and the device usage trace. And the power impact estimator computes an increase in power consumption of the mobile application based on the hardware usage statistics.

Communication device for predicting power consumption of mobile application, communication system including same, method of predicting power consumption of mobile application and method of providing predicted power consumption of mobile application, using same

Systems, methods, and computer program products customizing and delivering contextually relevant, artificially synthesized, voiced content that is targeted toward the individual user behaviors, viewing habits, experiences and preferences of each individual user accessing the content of a content provider. A network accessible profile service collects and analyzes collected user profile data and recommends contextually applicable voices based on the user's profile data. As user input to access voiced content or triggers voiced content maintained by a content provider, the voiced content being delivered to the user is a modified version comprising artificially synthesized human speech mimicking the recommended voice and delivering the dialogue of the voiced content, in a manner that imitates the sounds and speech patterns of the recommended voice.

Personalized custom synthetic speech

A computer-implemented method, according to one embodiment, includes: receiving a complex audio signal which includes an intended audio signal and at least one interfering audio signal. Moreover, the intended audio signal is a voice-based command originating from a user. Information which corresponds to the at least one interfering audio signal is also received. The received information is used to identify portions of the complex audio signal as being the at least one interfering audio signal. Furthermore, the identified portion of the complex audio signal is removed from the complex audio signal, and a remaining portion of the complex audio signal is output.

Filtering audio-based interference from voice commands using interference information

A refrigerator-implemented method for optimizing placement of food items in different compartments of a refrigerator, according to one embodiment, includes: receiving a request to store a first type of food item in the refrigerator, determining, using detection hardware, each type of food item already stored in one or more of the compartments of the refrigerator, comparing a characteristic of the first type of food item with a same type of characteristic for each type of food item already stored in the one or more of the compartments of the refrigerator, selecting a first of the compartments in which to store the first type of food item based at least in part on the comparison, and outputting, using an output device, a recommendation to store the first type of food item in the first compartment. Other systems, methods, and computer program products are described in additional embodiments.

Content and context aware regional cooling optimization for refrigerators

A computer-implemented method, according to one embodiment, includes: receiving a complex audio signal which includes an intended audio signal and at least one interfering audio signal. The complex audio signal is converted into text which represents a plurality of words included in the complex audio signal, and at least some of the text is identified as representing words which correspond to the at least one interfering audio signal. The identified text is discarded, and a remaining portion of the text is evaluated to determine whether the remaining portion of the text represents words which convey the voice-based command at an accuracy that is in a predetermined range. Furthermore, the remaining portion of the text is output in response to determining that the remaining portion of the text represents words which convey the voice-based command at an accuracy that is in the predetermined range.

Filtering audio-based interference from voice commands using natural language processing

A computer-implemented method according to one embodiment includes identifying, utilizing a hardware camera of a mobile device, vision code data from a plurality of additional devices that are within a field of view of the hardware camera of the mobile device at an actual orientation, identifying the actual orientation of the mobile device, calculating local observations at the mobile device, sending the local observations from the mobile device to a cloud-side service, receiving a tile at the mobile device from the cloud-side service, and outputting the tile at the mobile device.

Providing visualization data to a co-located plurality of mobile devices

A method enables object label persistence between subsequent images captured by a camera. One or more processors receive a first image, which is captured by an image sensor on a camera, and which includes a depiction of an object. The processor(s) generate a label for the object, and display the first image on a display. The processor(s) subsequently receive movement data that describes a movement of the camera after the image sensor on the camera captures the first image and before the image sensor on the camera captures a second image. The processor(s) receive the second image. The processor(s) display the second image on the display, and then detect a pixel shift between the first image and the second image as displayed on the display. The processor(s) then label the object with the label on the second image as displayed on the display.

Image labeling

A contemporary image is stored, is analyzed, and an associated data structure is generated. An image data structure may include image meta data, image classification data, image segment data, image caption data, or the like. The storage of the contemporary image is used as a triggering event and its data structure is used as an input to determine contextually relevant ancestral images. The contemporary image data structure is compared to respective data structures of ancestral images. A series of ancestral images are ranked based upon contextual similarly of their data structures to the contemporary image data structure. The most similar ancestral image is simultaneously displayed with the contemporary image. Interactions may be monitored, a prompt may be provided, a next ranked ancestral image may be displayed in place of the previous ancestral image, and/or a prompt to store a new contemporary image may be provided.

Juxtaposing contextually similar cross-generation images

A computer-implemented method according to one embodiment includes identifying a plurality of devices, determining a relative location of each of the plurality of devices, and assigning visualization data to each of the plurality of devices, based on the relative location of each of the plurality of devices, wherein determining the relative location of each of the plurality of devices includes performing relative localization on each of the plurality of devices by performing a constrained nonlinear optimization in order to reconstruct relative locations of each of the plurality of devices, using distributed observations received from the plurality of devices in order to determine a co-located group of the devices.

A computer-implemented method, according to one embodiment, includes: detecting a first type of food item added to a first compartment of a refrigerator, determining an ideal storage temperature associated with the type of food item added to the first compartment, adjusting an ambient storage temperature in the first compartment to substantially match the ideal storage temperature associated with the first type of food item, detecting a second type of food item added to a second compartment of the refrigerator, determining an ideal storage temperature associated with the second type of food item added to the second compartment, and adjusting an ambient storage temperature in the second compartment to substantially match the ideal storage temperature associated with the second type of food item. The ideal storage temperature associated with the first type of food item is different than the ideal storage temperature associated with the second type of food item.

Content and context aware microscopic cooling optimization for refrigerators

A method according to one embodiment includes identifying a plurality of mobile devices; sending a unique identifier to each of the plurality of mobile devices; receiving results of mobile vision sensing from each of the plurality of mobile devices, where the results of mobile vision sensing include, for each of the plurality of mobile devices, an identification of the unique identifier of other mobile devices of the plurality of mobile devices that are within a field of view of the mobile device; determining a relative location of each of the plurality of mobile devices, utilizing the results of mobile vision sensing; pixelating textual output, graphical output, or textual and graphical output into a matrix of tiles; assigning each tile to one of the plurality of mobile devices at a location matching a respective location within the matrix of tiles; and sending visualization data to the plurality of mobile devices.

Growing community of inventors

Yuseong-gu, South Korea

Chungkuk Yoo