Patents by Inventor Haichao Zhang
Haichao Zhang has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Publication number: 20220136760Abstract: A side-open left-right-door arbitrary-opening structure of a refrigerator comprises four hinge shafts are fixedly connected at left and right sides of a refrigerator body in pairs respectively, left and right sides of upper and lower end faces of a refrigerator door are all provided with an opening neck, each hinge shaft is correspondingly placed in one opening neck, four hinge shaft stoppers are slidably arranged on the upper and lower end faces of the refrigerator door in pairs respectively, one spring is arranged between two hinge shaft stoppers in each pair, tail ends of the hinge shaft stoppers on the upper and lower end faces of the refrigerator door extend into the opening necks at the corresponding sides, and the hinge shafts at the corresponding sides are positioned in the opening necks.Type: ApplicationFiled: January 14, 2022Publication date: May 5, 2022Inventors: Ping WU, Haichao ZHANG, Yanxi CHONG, Wenwen NIE, Qinan WU
-
Patent number: 11199355Abstract: Provided are a device (100) for self-adaptive regulation of air volume and a refrigerator (200) having the same.Type: GrantFiled: March 31, 2017Date of Patent: December 14, 2021Assignees: HEFEI HUALING CO., LTD., HEFEI MIDEA REFRIGERATOR CO., LTD., MIDEA GROUP CO., LTD.Inventors: Jun Li, Xiuhua Geng, Haichao Zhang
-
Patent number: 11117744Abstract: A user may place an item at an inventory location in an incorrect location, such as the wrong lane on the shelf. An untidy score is generated that is indicative of whether an item is where it is supposed to be. A correct item verification (CIV) score is determined that is indicative of how similar the returned item is to the type of item that is supposed to be stored at that location. An invalid item recognition (IIR) score is determined that is indicative of similarity of the returned item to a set of items associated with the user, that set excluding the item that is supposed to be stored at that location. The CIV and the IIR are combined to generate an untidy score. Based on the untidy score, someone may be dispatched to clean up the inventory location.Type: GrantFiled: December 8, 2015Date of Patent: September 14, 2021Assignee: AMAZON TECHNOLOGIES, INC.Inventors: Gerard Guy Medioni, Haichao Zhang, Saurabh Sehgal, Liefeng Bo
-
Patent number: 11074829Abstract: Described herein are systems and methods for interactive language acquisition with one-shot concept learning through a conversational environment. Supervised language learning is limited by the ability of capturing mainly the statistics of training data, and is hardly adaptive to new scenarios or flexible for acquiring new knowledge without inefficient retraining or catastrophic forgetting. In one or more embodiments, conversational interaction serves as a natural interface both for language learning and for novel knowledge acquisition. Embodiments of a joint imitation and reinforcement approach are disclosed for grounded language learning through interactive conversation. An agent trained with this approach is able to actively acquire information by asking questions about novel objects and use the just-learned knowledge in subsequent conversations in a one-shot fashion. Results compared with other methods verified the effectiveness of embodiments disclosed herein.Type: GrantFiled: April 12, 2018Date of Patent: July 27, 2021Assignee: Baidu USA LLCInventors: Haichao Zhang, Haonan Yu, Wei Xu
-
Publication number: 20210197089Abstract: A method for displaying a skin of a virtual character includes: displaying a matchmaking user interface (UI) for matching k user accounts online for gaming in the same virtual world, k being a positive integer; obtaining a customization parameter of a customized skin model of a target virtual character from a server when the target virtual character is present in virtual characters used by the k user accounts; loading n target skin component materials of the customized skin model according to the customization parameter, n being a positive integer; and displaying a gaming UI. The customized skin model is one of complete skin set models corresponding to the target virtual character and includes n skin components in a same style. The gaming UI includes the target virtual character located in the virtual world, the target virtual character wearing the customized skin model.Type: ApplicationFiled: March 16, 2021Publication date: July 1, 2021Inventors: Tianyu WANG, Zidong MO, Haichao ZHANG, Yongliang HUANG, Chuan LV, Xiangya FENG
-
Patent number: 11037025Abstract: Described herein are embodiments for an approach to improve the robustness of an object detector against adversarial attacks. Existing attacks for object detectors and the impacts of individual task component on model robustness are systematically analyzed from a multi-task view of object detection. In one or more embodiments, a multi-task learning perspective of object detection is introduced and an asymmetric role of task losses is identified. One or more embodiments of an adversarial training method for robust object detection are presented to leverage the multiple sources of attacks for improving the robustness of object detection models.Type: GrantFiled: May 30, 2019Date of Patent: June 15, 2021Assignee: Baidu USA LLCInventors: Haichao Zhang, Jianyu Wang
-
Publication number: 20210122030Abstract: A bionic robot is provided, which includes a body; a plurality of sets of wheeled leg devices arranged at intervals in a front-rear direction, each comprising two wheeled leg devices arranged symmetrically in a left-right direction, each comprising a leg assembly and a travel wheel, and a power output shaft connected to the travel wheel; and a suspension device disposed in the body and connected to at least two sets from the plurality of sets of wheeled leg devices. The at least two sets of wheeled leg devices are located at the foremost end and the backmost end respectively. The suspension device comprises a plurality of drive assemblies, each connected to a corresponding leg assembly, which each comprise: an electric cylinder being configured to drive a telescopic rod to extend or retract; a damper connected between the body and the leg assembly; and an elastic member fitted over the damper.Type: ApplicationFiled: October 29, 2020Publication date: April 29, 2021Inventors: Jianwei ZHAO, Hai ZHANG, Zhikang CHI, Haichao ZHANG, Chaoyi LIU, Xiaofei MA
-
Patent number: 10936910Abstract: Described herein are embodiments for joint adversarial training methods that incorporate both spatial transformation-based and pixel-value based attacks for improving image model robustness. Embodiments of a spatial transformation-based attack with an explicit notion of budgets are disclosed and embodiments of a practical methodology for efficient spatial attack generation are also disclosed. Furthermore, both pixel and spatial attacks are integrated into embodiments of a generation model and the complementary strengths of each other are leveraged for improving the overall model robustness. Extensive experimental results on several benchmark datasets compared with state-of-the-art methods verified the effectiveness of the presented method.Type: GrantFiled: April 30, 2019Date of Patent: March 2, 2021Assignee: Baidu USA LLCInventors: Haichao Zhang, Jianyu Wang
-
Patent number: 10903646Abstract: The present disclosure relates to electrostatic protection terminals. One example terminal includes a target interface, a protected circuit, a protection unit, a switch unit, and a switch control unit. The protected circuit is configured to suppress an electrostatic discharge (ESD) current or an electrical overstress (EOS) current. A first end of the protection unit is electrically connected to a first pin of the target interface. A second end of the protection unit is electrically connected to a second pin of the protected circuit. The first pin is any pin of the target interface. The second pin is a pin that is in the protected circuit and that needs to be electrically connected to the first pin. The switch unit is connected to the protection unit in parallel. The switch control unit is configured to control the switch unit to be open or closed.Type: GrantFiled: July 26, 2016Date of Patent: January 26, 2021Assignee: Huawei Technologies Co., Ltd.Inventors: Haichao Zhang, Ying Xiong, Zhi Xiao
-
Publication number: 20210012188Abstract: Described herein are embodiments for a feature-scattering-based adversarial training approach for improving model robustness against adversarial attacks. Conventional adversarial training approaches leverage a supervised scheme, either targeted or non-targeted in generating attacks for training, which typically suffer from issues such as label leaking as noted in recent works. Embodiments of the disclosed approach generate adversarial images for training through feature scattering in the latent space, which is unsupervised in nature and avoids label leaking. More importantly, the presented approaches generate perturbed images in a collaborative fashion, taking the inter-sample relationships into consideration. Extensive experiments on different datasets compared with state-of-the-art approaches demonstrate the effectiveness of the presented embodiments.Type: ApplicationFiled: July 9, 2019Publication date: January 14, 2021Applicant: Baidu USA LLCInventors: Haichao ZHANG, Jianyu WANG
-
Publication number: 20200410228Abstract: Described herein are embodiments for fast training of adversarially robust models against adversarial attacks. Embodiments for model training by perturbing both the image and the label, which may be referred to as Bilateral Adversarial Training (BAT), are disclosed. To generate the adversarial label, one or more closed-form heuristic solutions are derived. One-step targeted attack is used with the target label being the most confusing class. It is shown in various experiments that random start and the most confusing target attack effectively prevent the label leaking and gradient masking problem. Coupled with the adversarial label part, embodiments of presented models significantly improve the state-of-the-art results. Experiments on one or more computationally challenging dataset also demonstrate the effectiveness of the presented BAT method embodiments.Type: ApplicationFiled: June 28, 2019Publication date: December 31, 2020Applicant: Baidu USA LLCInventors: Jianyu WANG, Haichao ZHANG
-
Publication number: 20200380300Abstract: Described herein are embodiments for an approach to improve the robustness of an object detector against adversarial attacks. Existing attacks for object detectors and the impacts of individual task component on model robustness are systematically analyzed from a multi-task view of object detection. In one or more embodiments, a multi-task learning perspective of object detection is introduced and an asymmetric role of task losses is identified. One or more embodiments of an adversarial training method for robust object detection are presented to leverage the multiple sources of attacks for improving the robustness of object detection models.Type: ApplicationFiled: May 30, 2019Publication date: December 3, 2020Applicant: Baidu USA LLCInventors: Haichao ZHANG, Jianyu WANG
-
Publication number: 20200265271Abstract: Described herein are embodiments for joint adversarial training methods that incorporate both spatial transformation-based and pixel-value based attacks for improving image model robustness. Embodiments of a spatial transformation-based attack with an explicit notion of budgets are disclosed and embodiments of a practical methodology for efficient spatial attack generation are also disclosed. Furthermore, both pixel and spatial attacks are integrated into embodiments of a generation model and the complementary strengths of each other are leveraged for improving the overall model robustness. Extensive experimental results on several benchmark datasets compared with state-of-the-art methods verified the effectiveness of the presented method.Type: ApplicationFiled: April 30, 2019Publication date: August 20, 2020Applicant: Baidu USA LLCInventors: Haichao ZHANG, Jianyu WANG
-
Publication number: 20190318648Abstract: Described herein are systems and methods for interactive language acquisition with one-shot concept learning through a conversational environment. Supervised language learning is limited by the ability of capturing mainly the statistics of training data, and is hardly adaptive to new scenarios or flexible for acquiring new knowledge without inefficient retraining or catastrophic forgetting. In one or more embodiments, conversational interaction serves as a natural interface both for language learning and for novel knowledge acquisition. Embodiments of a joint imitation and reinforcement approach are disclosed for grounded language learning through interactive conversation. An agent trained with this approach is able to actively acquire information by asking questions about novel objects and use the just-learned knowledge in subsequent conversations in a one-shot fashion. Results compared with other methods verified the effectiveness of embodiments disclosed herein.Type: ApplicationFiled: April 12, 2018Publication date: October 17, 2019Applicant: Baidu USA LLCInventors: Haichao ZHANG, Haonan YU, Wei XU
-
Publication number: 20190310007Abstract: Provided are a device (100) for self-adaptive regulation of air volume and a refrigerator (200) having the same.Type: ApplicationFiled: March 31, 2017Publication date: October 10, 2019Inventors: Jun LI, Xiuhua GENG, Haichao ZHANG
-
Patent number: 10366166Abstract: Described herein are systems and methods for human-like language acquisition in a compositional framework to implement object recognition or navigation tasks. Embodiments include a method for a model to learn the input language in a grounded and compositional manner, such that after training the model is able to correctly execute zero-shot commands, which have either combination of words in the command never appeared before, and/or new object concepts learned from another task but never learned from navigation settings. In embodiments, a framework is trained end-to-end to learn simultaneously the visual representations of the environment, the syntax and semantics of the language, and outputs actions via an action module. In embodiments, the zero-shot learning capability of a framework results from its compositionality and modularity with parameter tying.Type: GrantFiled: September 7, 2017Date of Patent: July 30, 2019Assignee: Baidu USA LLCInventors: Haonan Yu, Haichao Zhang, Wei Xu
-
Publication number: 20190165570Abstract: The present disclosure relates to electrostatic protection terminals. One example terminal includes a target interface, a protected circuit, a protection unit, a switch unit, and a switch control unit. The protected circuit is configured to suppress an electrostatic discharge (ESD) current or an electrical overstress (EOS) current. A first end of the protection unit is electrically connected to a first pin of the target interface. A second end of the protection unit is electrically connected to a second pin of the protected circuit. The first pin is any pin of the target interface. The second pin is a pin that is in the protected circuit and that needs to be electrically connected to the first pin. The switch unit is connected to the protection unit in parallel. The switch control unit is configured to control the switch unit to be open or closed.Type: ApplicationFiled: July 26, 2016Publication date: May 30, 2019Inventors: Haichao ZHANG, Ying XIONG, Zhi XIAO
-
Publication number: 20190073353Abstract: Described herein are systems and methods for human-like language acquisition in a compositional framework to implement object recognition or navigation tasks. Embodiments include a method for a model to learn the input language in a grounded and compositional manner, such that after training the model is able to correctly execute zero-shot commands, which have either combination of words in the command never appeared before, and/or new object concepts learned from another task but never learned from navigation settings. In embodiments, a framework is trained end-to-end to learn simultaneously the visual representations of the environment, the syntax and semantics of the language, and outputs actions via an action module. In embodiments, the zero-shot learning capability of a framework results from its compositionality and modularity with parameter tying.Type: ApplicationFiled: September 7, 2017Publication date: March 7, 2019Applicant: Baidu USA LLCInventors: Haonan YU, Haichao ZHANG, Wei XU
-
Publication number: 20180342174Abstract: Described herein are systems and methods for grounded natural language learning in an interactive setting. In embodiments, during a learning process, an agent learns natural language by interacting with a teacher and learning from feedback, thus learning and improving language skills while taking part in the conversation. In embodiments, a model is used to incorporate both imitation and reinforcement by leveraging jointly sentence and reward feedback from the teacher. Various experiments are conducted to validate the effectiveness of a model embodiment.Type: ApplicationFiled: November 22, 2017Publication date: November 29, 2018Applicant: Baidu USA LLCInventors: Haichao ZHANG, Haonan YU, Wei XU
-
Publication number: 20050080102Abstract: The subject invention relates to a novel small molecule, referred to as alpha-(trichloromethyl)-4-Pyridineethanol (PETCM), as well as to uses thereof. PETCM was identified and isolated by high throughput screening and is an activator of caspase-3. Using PETCM in combination with biochemical fractionation, a novel pathway that regulates mitochondria-initiated caspase activation was also identified. This pathway comprises tumor suppressor PHAP proteins and oncoprotein prothymosin-alpha. PETCM relieves prothymosin-alpha inhibition and allows apoptosome to form at a physiological concentration of dATP.Type: ApplicationFiled: July 12, 2004Publication date: April 14, 2005Inventors: Saul Rosenberg, Shi Ng, James Kofron, Haichao Zhang, Xuejun Jiang, Xiaodong Wang