Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
Response to Arguments
Applicant’s arguments with respect to claim(s) 16-18 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
In re pages 6-8, with respect to claims 8 and 12, applicants present that Lee fails to teach the specific feature map data object, which is defined narrowly as a partial representation of an image corresponding to the output, it teaches away from the processing operation performed as in claim 8, where a confidence is determined for an inference and then a separate data object is received by the server system.
In response, the examiner respectfully disagrees. Lee in paragraph 38 teaches “the edge device 300 performs operations up to a previous layer to an exit point EP…, and then calculates inference confidence based on the operation result… If it is determined that the inference confidence is lower than a threshold, the edge device 300 transmits an intermediate operation results obtained by processing up to corresponding layer, to the cloud 400. Then the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result.” The “intermediate operation result” is an intermediate layer neural network activation, i.e. a feature map partially representing the input image, generated and conditionally transmitted upon the same confidence threshold determination Chew discloses. Ther feature map data object is therefore met by the intermediate operation results since its represents data of a “partial representation of an image”. There does not appear to be any issues with rejection of limitations presented in claims 8 and 12.
Additionally, examiner makes note of the examiner’s remarks in re page 7 of “In response to a determination that the confidence does not exceed the threshold, the feature map is then generated as a different data object ….”. In looking through the specification with specific reference to Figs. 4A and 4B and supporting disclosure, it appears that the feature maps are referred to as “generated” in the past tense, that is sent to the server. There does not appear to be clear support for the “then generate” the feature map upon the condition.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102 of this title, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 8-9, 12-13, 19 and 22-23 are rejected under 35 U.S.C. 103 as being unpatentable over Chew et al. (US 2018/0181868) in view of Lee et al. (US 2020/0311546).
Regarding claim 8, Chew teaches a surveillance system (Fig. 4 and paragraph 33) comprising:
an edge device (paragraphs 16-17, 20 and 33) that is configured to:
generate a first set of images of an environment to be surveilled (paragraphs 15-17, 20 and 33 teaches camera capturing images/video),
apply a first model to each image in the first set of images, so as to produce a first set of outputs (Figs. 1-2, analytics determiner (i.e. “model”) running on local client device, running analytics to generate an output/result),
However, while Chew is explicit towards using a surveillance system to perform camera and server based analysis of video frames, fails to teach, however, Lee teaches:
wherein the first set of outputs is representative of an inference made, by the first model, based on the set of images (Figs. 1-3 and paragraphs 4, 38 and 43 teaches wherein an initial edge device 300 primarily working in the imaging/realm, and the images/sensor data to the cloud device/server. Lee in paragraph 38 teaches “the edge device 300 performs operations up to a previous layer to an exit point EP…, and then calculates inference confidence based on the operation result… If it is determined that the inference confidence is lower than a threshold, the edge device 300 transmits an intermediate operation results obtained by processing up to corresponding layer, to the cloud 400. Then the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result.” The “intermediate operation result” is an intermediate layer neural network activation, i.e. a feature map partially representing the input image, generated and conditionally transmitted upon the same confidence threshold determination Chew discloses. The operation results, from the various number of layers of the CNN/DDNN partially represents the input image in that it encodes spatial feature information extracted from the image through convolution, while not preserving the raw pixel-level representation of the image), and
for each output in the first set of outputs, determine whether confidence in that output exceeds a threshold (Figs. 1-3 and paragraphs 4 and 38 teaches determining whether a calculated confidence level exceeds a threshold), and
generate a feature map that consists of one or more features partially representing an image corresponding to that output in response to a determination that the confidence does not exceed the threshold (Figs. 1-3 and paragraphs 4, 38 and 43 teaches wherein an initial edge device 300 primarily working in the imaging/realm, and the images/sensor data to the cloud device/server. Lee in paragraph 38 teaches “the edge device 300 performs operations up to a previous layer to an exit point EP…, and then calculates inference confidence based on the operation result… If it is determined that the inference confidence is lower than a threshold, the edge device 300 transmits an intermediate operation results obtained by processing up to corresponding layer, to the cloud 400. Then the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result.” The “intermediate operation result” is an intermediate layer neural network activation, i.e. a feature map partially representing the input image, generated and conditionally transmitted upon the same confidence threshold determination Chew discloses. The operation results, from the various number of layers of the CNN/DDNN partially represents the input image in that it encodes spatial feature information extracted from the image through convolution, while not preserving the raw pixel-level representation of the image); and
cause transmission of the feature map to a server system (Figs. 1-3 and paragraphs 4 and 38 teaches transmitting the feature map to the server/cloud); and
the server system that is configured to: receive a set of feature maps from the edge device (Figs. 1-3 and paragraphs 4, 38 and 43 teaches the cloud/server system is sent the feature maps); and
provide the feature map as input to a second model so as to produce a second set of outputs, wherein the second set of outputs is representative of an inference made, by the second model, based on the feature map (Fig. 4 and paragraphs 33-39 teaches wherein the cloud device/server runs its own analysis sequentially through secondary layers all the way up to a final layer. Each layer outputs an inference until a final output result for the inference is output as a result (“cloud exit”). The proposed combination further discloses a server system configured to receive the feature map and provide it as input to a second model so as to produce a second set of outputs representative of an inference made by the second model, based on the feature map: Lee paragraph 38: “the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to incorporate the teachings of Lee into the system of Chew because combining Lee and Chew requires only routine skill and produces predictable results. The modification involves applying Chew’s confidence threshold comparison mechanism – a discrete, well-defined software component comprising a confidence level comparator, and a data sender (see paragraphs 43 and Figs. 6 and 9) – to Lee’s existing confidence evaluation step. Lee already implements confidence comparison as a core architectural element (paragraph 38); the modification is the substitution of a user-configurable threshold value for a predetermined one. There is no teaching away in either reference since Chew does not criticize user-defined thresholds as inferior to predetermined ones, and Lee does not teach that its predetermined threshold is structurally essential. The combination produces the predictable result of a more flexible and deployable distributed DNN inference system. See KSR v. Teleflex Inc.
Regarding claim 9, Lee teaches the claimed wherein each feature map is provided as input to a middle layer of the second model (Fig. 2 and 6C, paragraphs 4, 38, 40, 58 teaches the claimed “intermediate layer” in the form of at least “conv” layer as part of the cloud’s model).
Regarding claim 12, Chew teaches a method performed by an edge device (Fig. 4 and paragraph 33) that generates samples while surveilling an environment, the method comprising:
applying a first model to the samples to produce outputs (Figs. 1-2, analytics determiner (i.e. “model”) running on local client device, running analytics to generate an output/result),
However, while Chew is explicit towards using a surveillance system to perform camera and server based analysis of video frames, fails to teach, however, Lee teaches:
wherein each output is representative of an inference made in relation to a corresponding sample (Figs. 1-3 and paragraphs 4 and 38 teaches determining an inference based on an initial model output in a first layer/model);
determining whether confidence in each of the outputs exceeds a threshold (Figs. 1-3 and paragraphs 4 and 38 teaches determining whether a calculated confidence level exceeds a threshold); and
for each output for which the confidence does not exceed the threshold (Figs. 1-3 and paragraphs 4 and 38 teaches determining whether a calculated confidence level exceeds a threshold, and if below the threshold, inference is transmitted to a cloud/server system),
causing transmission of a feature map that consists of one or more features partially representing the corresponding sample to a server system, wherein the server system provides the feature map as input to a second model (Figs. 1-3 and paragraphs 4, 38 and 43 teaches wherein an initial edge device 300 primarily working in the imaging/realm, and the images/sensor data to the cloud device/server. Lee in paragraph 38 teaches “the edge device 300 performs operations up to a previous layer to an exit point EP…, and then calculates inference confidence based on the operation result… If it is determined that the inference confidence is lower than a threshold, the edge device 300 transmits an intermediate operation results obtained by processing up to corresponding layer, to the cloud 400. Then the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result.” The “intermediate operation result” is an intermediate layer neural network activation, i.e. a feature map partially representing the input image, generated and conditionally transmitted upon the same confidence threshold determination Chew discloses. The operation results, from the various number of layers of the CNN/DDNN partially represents the input image in that it encodes spatial feature information extracted from the image through convolution, while not preserving the raw pixel-level representation of the image. The proposed combination further discloses a server system configured to receive the feature map and provide it as input to a second model so as to produce a second set of outputs representative of an inference made by the second model, based on the feature map: Lee paragraph 38: “the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to incorporate the teachings of Lee into the system of Chew because combining Lee and Chew requires only routine skill and produces predictable results. The modification involves applying Chew’s confidence threshold comparison mechanism – a discrete, well-defined software component comprising a confidence level comparator, and a data sender (see paragraphs 43 and Figs. 6 and 9) – to Lee’s existing confidence evaluation step. Lee already implements confidence comparison as a core architectural element (paragraph 38); the modification is the substitution of a user-configurable threshold value for a predetermined one. There is no teaching away in either reference since Chew does not criticize user-defined thresholds as inferior to predetermined ones, and Lee does not teach that its predetermined threshold is structurally essential. The combination produces the predictable result of a more flexible and deployable distributed DNN inference system. See KSR v. Teleflex Inc.
Regarding claim 13, Chew teaches the claimed wherein the edge device is a camera, and wherein the model is trained to detect instances of an object in images (paragraphs 16-17, 20 and 33 teaches a camera and paragraphs 2-3 teaches object detection).
Regarding claim 19, Chew teaches the claimed wherein the edge device is further configured to: for each output in the first set of outputs, indicate that output is an appropriate inference in a data structure in response to a determination that the confidence does exceed the threshold (Figs. 1-3 and paragraphs 4 and 38 teaches determining whether a calculated confidence level exceeds a threshold and the result is appropriate and is output as an “exit”).
Regarding claim 22, Chew and Lee teaches the claimed wherein the first model requires fewer computational resources than the second model to produce an output when applied to a given image (as discussed above, the first model run on Chew’s system is less robust than the cloud system of Lee). The prior motivation as discussed above is incorporated herein.
Regarding claim 23, Chew and Lee teaches the claimed wherein the threshold is programmed in memory of the camera (Chew: Fig. 1 and paragraph 25: threshold data is stored on client device. Lee: paragraphs 4 and 38 teaches the threshold being part of the system, which includes the edge device 300). The prior motivation as discussed above is incorporated herein.
Claims 16-18 are rejected under 35 U.S.C. 103 as being unpatentable over Chew et al. (US 2018/0181868) in view of Lee et al. (US 2020/0311546) and further in view of Matsubara et al., “Head Network Distillation: Splitting Distilled Deep Neural Networks for Resource-Constrained Edge Computing Systems,” IEEE Access, vol. 8, pp.212177-212193, Nov. 2020, doi:10.1109/ACCESS.2020.3039714.
Regarding claim 16, Chew teaches a method performed by a server system (Chew: Fig. 4 and paragraph 33), the method comprising:
generated by the edge device while surveilling an environment (Figs. 1-2, analytics determiner (i.e. “model”) running on local client device, running analytics to generate an output/result);
storing an indication of the inference in a data structure (paragraph 17).
However, while Chew fails to teach the remaining limitations, Lee also teaches a method performed by a server system (Figs. 1-3), also performing:
receiving a feature map generated by an edge device (paragraphs 15-17, 20 and 33 teaches camera capturing images/video), wherein the feature map consists of one or more features partially representing a sample (Figs. 1-3 and paragraphs 4, 38 and 43 teaches wherein an initial edge device 300 primarily working in the imaging/realm, and the images/sensor data to the cloud device/server. Lee in paragraph 38 teaches “the edge device 300 performs operations up to a previous layer to an exit point EP…, and then calculates inference confidence based on the operation result… If it is determined that the inference confidence is lower than a threshold, the edge device 300 transmits an intermediate operation results obtained by processing up to corresponding layer, to the cloud 400. Then the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result.” The “intermediate operation result” is an intermediate layer neural network activation, i.e. a feature map partially representing the input image, generated and conditionally transmitted upon the same confidence threshold determination Chew discloses. The operation results, from the various number of layers of the CNN/DDNN partially represents the input image in that it encodes spatial feature information extracted from the image through convolution, while not preserving the raw pixel-level representation of the image. Therefore, the feature map is received by the server device from the edge device);
providing the feature map to the layer of the second model as input, so as to produce an output that is representative of an inference made in relation to the sample (Fig. 2 and 6C, paragraphs 4, 38, 40, 58 teaches the claimed “intermediate layer” in the form of at least “conv” layer as part of the cloud’s model that receives the operation result from an edge device 300. Then the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result.” The “intermediate operation result” is an intermediate layer neural network activation, i.e. a feature map partially representing the input image, generated and conditionally transmitted upon the same confidence threshold determination Chew discloses. The operation results, from the various number of layers of the CNN/DDNN partially represents the input image in that it encodes spatial feature information extracted from the image through convolution, while not preserving the raw pixel-level representation of the image), so as to produce a second set of outputs (Fig. 4 and paragraphs 33-39 teaches wherein the cloud device/server runs its own analysis sequentially through secondary layers all the way up to a final layer. Each layer outputs an inference until a final output result for the inference is output as a result (“cloud exit”). The proposed combination further discloses a server system configured to receive the feature map and provide it as input to a second model so as to produce a second set of outputs representative of an inference made by the second model, based on the feature map: Lee paragraph 38: “the cloud 400 performs operations from a subsequent layer to the exit point EP up to a last layer, and infers and then outputs a final result”)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to incorporate the teachings of Lee into the system of Chew because combining Lee and Chew requires only routine skill and produces predictable results. The modification involves applying Chew’s confidence threshold comparison mechanism – a discrete, well-defined software component comprising a confidence level comparator, and a data sender (see paragraphs 43 and Figs. 6 and 9) – to Lee’s existing confidence evaluation step. Lee already implements confidence comparison as a core architectural element (paragraph 38); the modification is the substitution of a user-configurable threshold value for a predetermined one. There is no teaching away in either reference since Chew does not criticize user-defined thresholds as inferior to predetermined ones, and Lee does not teach that its predetermined threshold is structurally essential. The combination produces the predictable result of a more flexible and deployable distributed DNN inference system. See KSR v. Teleflex Inc.
However, while Chew and Lee teaches the claimed as discussed above, fails to teach: generating, by an adaptation module, an adapted feature map, wherein the adapted feature map is a modified version of the feature map adapted to an output of a layer of a second model;
Matsubara remedies this deficiency, Matsubara discloses a split computing architecture in which a lightweight “head” network, and architecture distinct from, and substantially smaller than, the target network, is deployed on a resource constrained edge/mobile device and produces a compressed intermediate feature from input data. The compressed feature is transmitted to an edge server. At the server, a decode reconstructs the received feature into a tensor that approximates the output that a specific, targeted intermediate layer of a separately and previously trained, larger (”teacher/original”) deep neural network would itself have produced. The decoder is trained such that its reconstructed output is compatible with, and adapted to, that teach layer’s expected input. The “teacher” network’s own remaining (“tail”) layers, unmodified, then process the reconstructed tensor through to a final inference (e.g., classification or object detection). In particular to the claim language, Matsubara’s decoder is an “adaptation module”; its reconstructed tensor is “an adapted feature map” that is “a modified version of the feature map”, i.e., of the compressed feature received from the head network, “adapted to an output of a layer of a second model,” since it is trained specifically to approximate the intermediate layer output of the separately pretrained teacher network. Supplying that reconstructed tensor to the teacher’s next layer discloses “providing the feature map to an intermediary layer of the second model as input, so as to produce an output that is representative of an inference made in relation to the sample.” See pages 212178 (col. 2), 212179 (col. 2), 212184 (section IV.C, Eq. 4 and IV.D, 212186 (section V.A.2).
A person of ordinary skill in the art at the time of the filling of the instant application would have had a reason to modify the Chew/Lee combination to incorporate Matsubara’s decoder/adaptation step. Lee’s architecture requires the edge devise to be provisioned with, and kept in sync with, a literal partition of the exact same trained model deployed at the cloud, the edge and cloud portions are two pieces of one network. That constrains the edge devices model to whatever size and architecture the deployed cloud model happens to use at the chosen partition point. A POSITA seeking to reduce the edge devices computational and memory burden further than a mere partition allows, for instance, to accommodate a smaller or cheaper class of edge hardware, or to allow the server side to run an already deployed, independently maintained and more accurate model without reprovisioning every edge device with a matching partition of that specific model, would have looked to Matsubara, which addresses precisely that same problem. The problem being that it trains a separate, smaller head network for the constrained device and supplies a decoder that reconciles the mismatch between that head’s output and the intermediate representation expected by an existing, independently pretrained larger network. This is a combination of known elements, Lee’s confidence driven edge cloud feature transmission architecture and Matsubara's feature reconstruction adaptation technique, each performing the same function it performs in isolation, combined by known methods to yield the predictable result of enabling a small, independently architected edge model to interoperate with a larger previously trained server-side model. See KSR Int’l Col. V. Teleflex Inc., 550 U.S. 398, 416, 420-421 (2007). Matsubara’s decoder is not being imported as a whole into a foreign context, it is applied to the same technical problem (device-edge split DNN inference under bandwidth and edge compute constraints) that motivates both Chew and Lee, using the same type of intermediate feature data.
Regarding claim 17, Chew and Lee teaches the claimed wherein the data structure is maintained in memory of the server system (Chew: paragraph 17 and Lee: paragraph 38-40).
Regarding claim 18, Chew teaches the claimed wherein the sample is representative of a digital image of the environment (paragraphs 15-17, 20 and 33 teaches camera capturing images/video).
Claims 10-11 are rejected under 35 U.S.C. 103 as being unpatentable over Chew et al. (US 2018/0181868) in view of Lee et al. (US 2020/0311546) and further in view of Yang et al. (US 2017/0076195).
Regarding claim 10, Chew and Lee teaches the claimed as discussed in claim 8, wherein Chew’s system is able to perform classification/object detection operations and Lee’s DNN also is processing images to determine inferences, but fails to teach, but Yang teaches the claimed wherein the first and second models are classification models (paragraphs 31 and 33, classifier functions).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to incorporate the teachings of Yang such that the generic functions of Chew and Lee would also allow for the use in classification functions in both the first and second models as taught by Yang because said incorporation allows for the benefit of enhancing user experience when interacting with such devices (paragraphs 1-4).
Regarding claim 11, Chew and Lee teaches the claimed as discussed in claim 8, wherein Chew’s system is able to perform classification/object detection operations and Lee’s DNN also is processing images to determine inferences, but Yang teaches the claimed wherein the first and second models are object detection models (paragraphs 36-37, object detection models).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to incorporate the teachings of Yang such that the generic functions of Chew and Lee would also allow for the use in classification functions in both the first and second models as taught by Yang because said incorporation allows for the benefit of enhancing user experience when interacting with such devices (paragraphs 1-4).
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Chew et al. (US 2018/0181868) in view of Lee et al. (US 2020/0311546) in view of Nitz et al. (US 2015/0213371).
Regarding claim 15, Chew teaches the claimed as discussed in claim 1 above, however fails to but Nitz wherein said applying, said determining, and said causing are performed in real time as the samples are generated by the edge device (paragraphs 0043, 0075 and 0076).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to incorporate the teachings of Nitz into the system of Chew and Lee because said incorporation allows for the benefit of determining an interference context based on real-time input (abstract).
Claims 20-21 are rejected under 35 U.S.C. 103 as being unpatentable over Chew et al. (US 2018/0181868) in view of Lee et al. (US 2020/0311546) and further in view of Takeuchi et al. (US 2010/0295944).
Regarding claim 20, Chew in its proposed combination with Lee fails to teach, however, Takeuchi teaches the claimed wherein the server system is further configured to: cause transmission of the second set of outputs to the edge device (Figs. 5-7 and paragraph 76 teaches wherein analysis from server 130 is sent back to the camera 100); and wherein the camera is further configured to: establish that an activity or an object of interest is included in at least one image in the first set of images (Figs. 5-7 and paragraphs 62-76, metadata combing section combines initial metadata generated by the camera 100 with metadata from the analysis server 130. Paragraph 79-81 teaches wherein event information is generated by the rules engine section within the camera) based on an analysis of (i) confident outputs in the first set of outputs and (ii) the second set of outputs (Figs. 5-7 and paragraphs 76-79, metadata combing section combines initial metadata generated by the camera 100 with metadata from the analysis server 130), and cause a notification that specifies the activity or the object of interest to be presented by a computer program executing on a mediatory device (Fig. 5, metadata and event data is sent to a central server 10 and paragraph 80-81 and 121 teaches wherein the an alert is generated on server 10 to alert a user of an event).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to incorporate the teachings of Takeuchi into the system of Chew and Lee such that the results of the analysis is sent back to the camera in Chew because such an incorporation allows for the benefit of sharing resources to assist sink/edge devices with additional resources when required (paragraphs 7-19).
Regarding claim 21, Chew and Lee teaches the claimed as discussed in claim 1 above, however fails to teach, but Takeuchi teaches wherein the server system is further configured to: cause transmission of the second set of outputs to a computer program executing on a mediatory device (Figs. 5 and 12, wherein first and second outputs (metadata A1 and metadata B3, event is sent to a central server 10); and wherein the camera is further configured to: cause transmission of confident outputs in the first set of outputs to the computer program executing on the mediatory device (Figs. 5 and 12, wherein first and second outputs (metadata A1 and metadata B3, event is sent to a central server 10). The prior motivation as discussed above is incorporated herein.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GELEK W TOPGYAL whose telephone number is (571)272-8891. The examiner can normally be reached M-F (9:30-6 PST).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, William Vaughn can be reached at 571-272-3922. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/GELEK W TOPGYAL/Primary Examiner, Art Unit 2481
Prosecution Insights