Description:FORM 2

THE PATENTS ACT, 1970
(39 of 1970)
&
THE PATENT RULES, 2003

COMPLETE SPECIFICATION
(See Section 10 and Rule 13)

Title of invention:

METHOD AND SYSTEM FOR DOMAIN ADAPTATION OF ELECTRO-OPTICAL DATA AND SYNTHETIC-APERTURE RADAR (SAR) DATA

Applicant

Tata Consultancy Services Limited
A company Incorporated in India under the Companies Act, 1956
Having address:
Nirmal Building, 9th floor,
Nariman point, Mumbai 400021,
Maharashtra, India

Preamble to the description:
The following specification particularly describes the invention and the manner in which it is to be performed.
TECHNICAL FIELD
[001] The disclosure herein generally relates to domain adaptation, and, more particularly, to a method and system for domain adaptation of electro-optical and Synthetic-aperture radar (SAR) data.

BACKGROUND

[002] Off late, due to large availability of satellite imagery, large volume of data is captured every day. Such huge data corpus is used to train machine learning models for various remote sensing tasks. Electro-optical (e.g. Multi-spectral (MS)) and Synthetic-aperture radar (SAR) are popular imaging techniques. As MS and SAR data differ in terms of captured sensor, there exists a significant domain gap between the two modalities. Hence, they can be considered as two different domains. Thus, any machine learning model trained on one domain cannot be tested another domain. In order to address the domain shift, a large amount of annotated dataset is required for both domains. Unfortunately, obtaining annotated datasets for both domains is a tedious task. However, obtaining annotated ground truth to train the model in a supervised setting is challenging, time consuming and requires domain expertise.

SUMMARY
[003] Embodiments of the present disclosure present technological improvements as solutions to one or more of the above-mentioned technical problems recognized by the inventors in conventional systems. For example, in one embodiment, a system for domain adaptation is provided. The system includes an electro-optical encoder implemented by one or more hardware processors, a pretrained data model implemented by the one or more hardware processors, and a Synthetic-aperture radar (SAR) encoder implemented by the one or more hardware processors. The domain adaptation includes the following steps. Initially, the electro-optical encoder and the pretrained data model are trained by performing LULC classification of a plurality of electro-optical data, to generate a LULC classification map, wherein the LULC classification map includes a plurality of electro-optical encoder features. Further, the SAR encoder is trained using SAR data, freezing weights of the electro-optical encoder and the pretrained data model, wherein training the SAR encoder includes the following steps. In this process, a feature map of the SAR data is processed using a plurality of Neural Weighted Averaging (NeWA) modules of the SAR encoder, wherein each of the plurality of NeWA modules generates a weight value corresponding to the feature map. Further, a consolidated feature map is generated at each of the plurality of NeWA modules by applying a plurality of the weight value generated by the plurality of NeWA modules on the feature map of the SAR data. Further, an output feature map is generated by concatenating a plurality of the consolidated feature maps weight value generated by the plurality of NeWA modules, at each of the plurality of NeWA modules, wherein, the output feature map forms a SAR data model. After training the SAR encoder, a Feature Domain Adaptation (FDA) loss between the electro-optical encoder features and the SAR data model is computed. Further, the computed FDA loss is compared with a threshold of FDA loss. Further, the SAR encoder is retrained till the computed FDA loss is below a threshold, if the computed FDA loss is exceeding the threshold of FDA loss.
[004] In another aspect, a processor implemented method of domain adaptation is provided. In this method, initially an electro-optical encoder and a pretrained data model of a domain adaptation system are trained by performing LULC classification of a plurality of electro-optical data, to generate a LULC classification map, wherein the LULC classification map includes a plurality of electro-optical encoder features. Further, an SAR encoder is trained using SAR data, freezing weights of the electro-optical encoder and the pretrained data model, wherein training the SAR encoder includes the following steps. In this process, a feature map of the SAR data is processed using a plurality of Neural Weighted Averaging (NeWA) modules of the SAR encoder, wherein each of the plurality of NeWA modules generates a weight value corresponding to the feature map. Further, a consolidated feature map is generated at each of the plurality of NeWA modules by applying a plurality of the weight value generated by the plurality of NeWA modules on the feature map of the SAR data. Further, an output feature map is generated by concatenating a plurality of the consolidated feature maps weight value generated by the plurality of NeWA modules, at each of the plurality of NeWA modules, wherein, the output feature map forms a SAR data model. After training the SAR encoder, a Feature Domain Adaptation (FDA) loss between the electro-optical encoder features and the SAR data model is computed. Further, the computed FDA loss is compared with a threshold of FDA loss. Further, the SAR encoder is retrained till the computed FDA loss is below a threshold, if the computed FDA loss is exceeding the threshold of FDA loss.
[005] In yet another aspect, a non-transitory computer readable medium for domain adaptation is provided. The non-transitory computer readable medium includes a plurality of instructions which when executed, causes the following steps. Initially an electro-optical encoder and a pretrained data model of a domain adaptation system are trained by performing LULC classification of a plurality of electro-optical data, to generate a LULC classification map, wherein the LULC classification map includes a plurality of electro-optical encoder features. Further, an SAR encoder is trained using SAR data, freezing weights of the electro-optical encoder and the pretrained data model, wherein training the SAR encoder includes the following steps. In this process, a feature map of the SAR data is processed using a plurality of Neural Weighted Averaging (NeWA) modules of the SAR encoder, wherein each of the plurality of NeWA modules generates a weight value corresponding to the feature map. Further, a consolidated feature map is generated at each of the plurality of NeWA modules by applying a plurality of the weight value generated by the plurality of NeWA modules on the feature map of the SAR data. Further, an output feature map is generated by concatenating a plurality of the consolidated feature maps weight value generated by the plurality of NeWA modules, at each of the plurality of NeWA modules, wherein, the output feature map forms a SAR data model. After training the SAR encoder, a Feature Domain Adaptation (FDA) loss between the electro-optical encoder features and the SAR data model is computed. Further, the computed FDA loss is compared with a threshold of FDA loss. Further, the SAR encoder is retrained till the computed FDA loss is below a threshold, if the computed FDA loss is exceeding the threshold of FDA loss.
[006] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.

BRIEF DESCRIPTION OF THE DRAWINGS
[007] The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate exemplary embodiments and, together with the description, serve to explain the disclosed principles:
[008] FIG. 1 illustrates an exemplary system for domain adaptation of electro-optical data and Synthetic-Aperture Radar (SAR) data, according to some embodiments of the present disclosure.
[009] FIG. 2 is a functional block diagram of the system of FIG. 1, according to some embodiments of the present disclosure.
[010] FIG. 3 is a functional block diagram of a Neural Weighted Averaging (NeWA) module of the system of FIG. 2, according to some embodiments of the present disclosure.
[011] FIGS. 4A and 4B (collectively referred to as FIG. 4) is a flow diagram depicting steps involved in the process of domain adaptation being performed by the system of FIG. 1, in accordance with some embodiments of the present disclosure.
[012] FIG. 5 is a flow diagram depicting steps involved in the process of generating weight value corresponding to a feature map by each of a plurality of NeWA modules of the system of FIG. 2, in accordance with some embodiments of the present disclosure.

DETAILED DESCRIPTION OF EMBODIMENTS
[013] Exemplary embodiments are described with reference to the accompanying drawings. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. Wherever convenient, the same reference numbers are used throughout the drawings to refer to the same or like parts. While examples and features of disclosed principles are described herein, modifications, adaptations, and other implementations are possible without departing from the scope of the disclosed embodiments.
[014] Electro-optical imaging (such as Multi-spectral (MS) imaging) and Synthetic-aperture radar (SAR) are popular imaging techniques. As electro-optical data and SAR data differ in terms of captured sensor, there exists a significant domain gap between the two modalities. Thus, any machine learning model trained on one domain cannot be tested another domain. In order to address the domain shift, a large amount of annotated dataset is required for both domains. Unfortunately, obtaining annotated datasets for both domains is a tedious task. However, obtaining annotated ground truth to train the model in a supervised setting is challenging, time consuming and requires domain expertise.
[015] System and method disclosed in the embodiments herein provide a domain adaptation approach for the electro-optical and SAR data. In this method, initially an electro-optical encoder and a pretrained data model are trained using electro-optical data to generate a LULC classification map, wherein the LULC classification map includes a plurality of electro-optical features. Further, an SAR encoder is trained by freezing weights of the electro-optical encoder and the pretrained data model, wherein, during this process a plurality of Neural Weighted Averaging (NeWA) modules of the SAR encoder generate weight vector specific to an input feature map, and in turn generates an SAR data model. After training the SAR encoder, FDA (Feature Domain Adaptation) feature loss which represents extent of difference between the electro-optical encoder features and the SAR data model, is computed. The computed FDA feature loss is then compared with a threshold of FDA feature loss, and if the computed FDA feature loss exceeds the threshold of FDA feature loss, then the SAR encoder is retrained, till the computed FDA loss is below the threshold of FDA loss.
[016] Referring now to the drawings, and more particularly to FIG. 1 through FIG. 5, where similar reference characters denote corresponding features consistently throughout the figures, there are shown preferred embodiments and these embodiments are described in the context of the following exemplary system and/or method.
[017] FIG. 1 illustrates an exemplary system for domain adaptation of electro-optical data and Synthetic-Aperture Radar (SAR) data, according to some embodiments of the present disclosure. The system 100 includes or is otherwise in communication with hardware processors 102, at least one memory such as a memory 104, an I/O interface 112. The hardware processors 102, memory 104, and the Input /Output (I/O) interface 112 may be coupled by a system bus such as a system bus 108 or a similar mechanism. In an embodiment, the hardware processors 102 can be one or more hardware processors.
[018] The I/O interface 112 may include a variety of software and hardware interfaces, for example, a web interface, a graphical user interface, and the like. The I/O interface 112 may include a variety of software and hardware interfaces, for example, interfaces for peripheral device(s), such as a keyboard, a mouse, an external memory, a printer and the like. Further, the I/O interface 112 may enable the system 100 to communicate with other devices, such as web servers, and external databases.
[019] The I/O interface 112 can facilitate multiple communications within a wide variety of networks and protocol types, including wired networks, for example, local area network (LAN), cable, etc., and wireless networks, such as Wireless LAN (WLAN), cellular, or satellite. For the purpose, the I/O interface 112 may include one or more ports for connecting several computing systems with one another or to another server computer. The I/O interface 112 may include one or more ports for connecting several devices to one another or to another server.
[020] The one or more hardware processors 102 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, node machines, logic circuitries, and/or any devices that manipulate signals based on operational instructions. Among other capabilities, the one or more hardware processors 102 is configured to fetch and execute computer-readable instructions stored in the memory 104.
[021] The memory 104 may include any computer-readable medium known in the art including, for example, volatile memory, such as static random-access memory (SRAM) and dynamic random-access memory (DRAM), and/or non-volatile memory, such as read only memory (ROM), erasable programmable ROM, flash memories, hard disks, optical disks, and magnetic tapes. In an embodiment, the memory 104 includes a plurality of modules 106.
[022] The plurality of modules 106 include programs or coded instructions that supplement applications or functions performed by the system 100 for executing different steps involved in the process of domain adaptation, being performed by the system 100. The plurality of modules 106, amongst other things, can include routines, programs, objects, components, and data structures, which performs particular tasks or implement particular abstract data types. The plurality of modules 106 may also be used as, signal processor(s), node machine(s), logic circuitries, and/or any other device or component that manipulates signals based on operational instructions. Further, the plurality of modules 106 can be used by hardware, by computer-readable instructions executed by the one or more hardware processors 102, or by a combination thereof. The plurality of modules 106 can include various sub-modules (not shown). The plurality of modules 106 may include computer-readable instructions that supplement applications or functions performed by the system 100 for the domain adaptation.
[023] The data repository (or repository) 110 may include a plurality of abstracted piece of code for refinement and data that is processed, received, or generated as a result of the execution of the plurality of modules in the module(s) 106. The plurality of modules 106 includes an electro-optical encoder 201, a pretrained data model 202, and a Synthetic-aperture radar (SAR) encoder 203.
[024] Although the data repository 110 is shown internal to the system 100, it will be noted that, in alternate embodiments, the data repository 110 can also be implemented external to the system 100, where the data repository 110 may be stored within a database (repository 110) communicatively coupled to the system 100. The data contained within such external database may be periodically updated. For example, new data may be added into the database (not shown in FIG. 1) and/or existing data may be modified and/or non-useful data may be deleted from the database. In one example, the data may be stored in an external system, such as a Lightweight Directory Access Protocol (LDAP) directory and a Relational Database Management System (RDBMS). Functions of the components of the system 100 are now explained with reference to the functional implementation depicted in FIGS. 2 and 3, and steps in flow diagrams in FIG. 4 and FIG. 5.
[025] FIG. 2 is a functional block diagram of the system of FIG. 1, according to some embodiments of the present disclosure.
[026] In the example implementation as in FIG. 2, the system 100 includes the electro-optical encoder 201, the pretrained data model 202, and the Synthetic-aperture radar (SAR) encoder 203. It is to be noted that for explanation purpose, electro-optical data considered is a Multi Scale (MS data), and hence the domain adaptation is explained with reference to MS-SAR domain adaptation example. For the same reason, the electro-optical encoder is alternately referred to as MS encoder 201, and the electro-optical data considered is MS data. Person skilled in the art would appreciate that the foregoing description would not limit scope of embodiments disclosed herein to MS-SAR domain adaptation, and that the approach disclosed in the embodiments herein is applicable to any electro-optical data to SAR domain adaptation.
[027] The MS encoder 201, the pretrained data model 202, and the Synthetic-aperture radar (SAR) encoder 203 are implemented by the one or more hardware processors 102. Functions of the components of the system 100 are explained with reference to steps in method 400 in FIG. 4. In an embodiment, the system 100 comprises one or more data storage devices or the memory 104 operatively coupled to the processor(s) 102 and is configured to store instructions for execution of steps of the method 400 by the processor(s) or one or more hardware processors 102. The steps of the method 400 of the present disclosure will now be explained with reference to the steps of flow diagrams in FIG. 4 and FIG. 5. Although process steps, method steps, techniques or the like may be described in a sequential order, such processes, methods, and techniques may be configured to work in alternate orders. In other words, any sequence or order of steps that may be described does not necessarily indicate a requirement that the steps to be performed in that order. The steps of processes described herein may be performed in any order practical. Further, some steps may be performed simultaneously.
[028] At step 402 of method 400 in FIG. 4, the system 100 trains the MS encoder 201 and the pretrained data 202 by performing Land Use Land Cover (LULC) classification of a plurality of MS data, to generate a LULC classification map. The LULC classification map includes a plurality of MS encoder features. In an embodiment, the pretrained data model maybe of any suitable type. For example, a DeepLabv3 model is used as the pretrained data model. The FIG. 2 depicts the stage of MS encoder training as a pre-training stage, and at this stage a Categorical Cross Entropy (CCE) loss maybe calculated for determining efficiency of the MS encoder training. Further, at step 404 of the method 400, the system 100 trains the SAR encoder 203 using SAR data, freezing weights of the MS encoder 201 and the pretrained data model 202. Various steps involved in the process of training the SAR encoder 203 are depicted in steps 404a through 404c. At step 404a, a feature map of the SAR data is processed using a plurality of Neural Weighted Averaging (NeWA) modules of the SAR encoder 203, wherein each of the plurality of NeWA modules works based on a fuzzy logic, and generates a weight value corresponding to the feature map. In an embodiment, the plurality of NeWA modules are configured to function such that the weight value is specific for the input SAR feature map. Arrangement of the NeWA modules to process input feature map of the SAR data is depicted in FIG. 3. As in FIG. 3, multiple NeWA modules are arranged in parallel, such that all the NeWA modules processes the input feature map simultaneously to generate respective weight values. In this process, which is depicted in method 500 in FIG. 5, each of the NeWA modules obtains a focused feature map at step 502 of the method 500, by performing feature extraction from the feature map, at a first, a second, and a third convolution layer of the NeWA module, wherein the focused feature map is a distilled knowledge of the feature map that is input to the NeWA module. Further, at step 504 of the method 500, a feature vector corresponding to the focused feature map is generated by each of the NeWA modules, wherein the feature vector is a 1-Dimensional weight vector corresponding to the feature map that is input to the NeWA module. Further, a consolidated feature map is generated at each of the plurality of NeWA modules by applying a plurality of the weight value generated by the plurality of NeWA modules on the feature map of the SAR data, at step 404b. Further, an output feature map is generated by concatenating a plurality of the consolidated feature maps weight value generated by the plurality of NeWA modules, at each of the plurality of NeWA modules, at step 404c, wherein, the output feature map forms a SAR data model.
[029] Further, at step 406 of the method 400, the system 100 computes a Feature Domain Adaptation (FDA) loss between the MS encoder features and the SAR data model. Value of the FDA loss represents extent to which the MS data and the SAR data are in synchronization, which is crucial to achieve the domain adaptation. Further, at step 408 of the method 400, the system 100 compares the computed FDA loss with a threshold of FDA loss. In an embodiment, value of the threshold of FDA loss maybe empirically determined, and maybe pre-configured with the system 100. In another embodiment, the value of the threshold of FDA loss can be reconfigured as per requirements. The domain adaptation for the MS data and the SAR data is considered to have been not achieved if the measured FDA loss exceeds the threshold of FDA loss. Hence, if the measured FDA loss value exceeds the threshold of FDA loss value, then at step 410 of the method 400, the system 100 retrains the SAR encoder till the computed FDA loss is below the threshold of FDA loss. The system 100 may terminate the training/retraining process if the measured FDA loss value is below the threshold of FDA loss, which indicates that loss is at an acceptable rate, and that the MS data and the SAR data are in synchronization, and hence the domain adaptation is successful.
[030] The written description describes the subject matter herein to enable any person skilled in the art to make and use the embodiments. The scope of the subject matter embodiments is defined by the claims and may include other modifications that occur to those skilled in the art. Such other modifications are intended to be within the scope of the claims if they have similar elements that do not differ from the literal language of the claims or if they include equivalent elements with insubstantial differences from the literal language of the claims.
[031] The embodiments of present disclosure herein address unresolved problem of MS and SAR domain adaptation. The embodiment, thus provides a method and system for domain adaptation of MS and SAR data. Moreover, the embodiments herein further provide a mechanism for customizing weight values used to process the data during the domain adaptation, by providing a plurality of Neural weighted Averaging (NeWA) modules in an SAR encoder used.
[032] It is to be understood that the scope of the protection is extended to such a program and in addition to a computer-readable means having a message therein; such computer-readable storage means contain program-code means for implementation of one or more steps of the method, when the program runs on a server or mobile device or any suitable programmable device. The hardware device can be any kind of device which can be programmed including e.g., any kind of computer like a server or a personal computer, or the like, or any combination thereof. The device may also include means which could be e.g., hardware means like e.g., an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a combination of hardware and software means, e.g., an ASIC and an FPGA, or at least one microprocessor and at least one memory with software processing components located therein. Thus, the means can include both hardware means and software means. The method embodiments described herein could be implemented in hardware and software. The device may also include software means. Alternatively, the embodiments may be implemented on different hardware devices, e.g., using a plurality of CPUs.
[033] The embodiments herein can comprise hardware and software elements. The embodiments that are implemented in software include but are not limited to, firmware, resident software, microcode, etc. The functions performed by various components described herein may be implemented in other components or combinations of other components. For the purposes of this description, a computer-usable or computer readable medium can be any apparatus that can comprise, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
[034] The illustrated steps are set out to explain the exemplary embodiments shown, and it should be anticipated that ongoing technological development will change the manner in which particular functions are performed. These examples are presented herein for purposes of illustration, and not limitation. Further, the boundaries of the functional building blocks have been arbitrarily defined herein for the convenience of the description. Alternative boundaries can be defined so long as the specified functions and relationships thereof are appropriately performed. Alternatives (including equivalents, extensions, variations, deviations, etc., of those described herein) will be apparent to persons skilled in the relevant art(s) based on the teachings contained herein. Such alternatives fall within the scope of the disclosed embodiments. Also, the words “comprising,” “having,” “containing,” and “including,” and other similar forms are intended to be equivalent in meaning and be open ended in that an item or items following any one of these words is not meant to be an exhaustive listing of such item or items, or meant to be limited to only the listed item or items. It must also be noted that as used herein and in the appended claims, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.
[035] Furthermore, one or more computer-readable storage media may be utilized in implementing embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory on which information or data readable by a processor may be stored. Thus, a computer-readable storage medium may store instructions for execution by one or more processors, including instructions for causing the processor(s) to perform steps or stages consistent with the embodiments described herein. The term “computer-readable medium” should be understood to include tangible items and exclude carrier waves and transient signals, i.e., be non-transitory. Examples include random access memory (RAM), read-only memory (ROM), volatile memory, nonvolatile memory, hard drives, CD ROMs, DVDs, flash drives, disks, and any other known physical storage media.
[036] It is intended that the disclosure and examples be considered as exemplary only, with a true scope of disclosed embodiments being indicated by the following claims.
, Claims:We Claim:

1. A system for domain adaptation, comprising:
one or more hardware processors (102);
a communication interface (112); and
a memory (104) storing a plurality of instructions, wherein the plurality of instructions stored in the memory cause the system to perform domain adaptation, wherein the domain adaptation comprising:
training an electro-optical encoder (201) and a pretrained data model (202) comprised in a plurality of modules in the memory, by performing Land Use Land Cover (LULC) classification of a plurality of electro-optical data, to generate a LULC classification map, wherein the LULC classification map comprises a plurality of electro-optical encoder features;
training a SAR encoder (203) comprised in the plurality of modules in the memory, using SAR data, freezing weights of the electro-optical encoder and the pretrained data model, comprising:
processing a feature map of the SAR data using a plurality of Neural Weighted Averaging (NeWA) modules of the SAR encoder, wherein each of the plurality of NeWA modules generates a weight value corresponding to the feature map;
generating a consolidated feature map at each of the plurality of NeWA modules by applying a plurality of the weight value generated by the plurality of NeWA modules on the feature map of the SAR data; and
generating an output feature map by concatenating a plurality of the consolidated feature maps weight value generated by the plurality of NeWA modules, at each of the plurality of NeWA modules, wherein,
the output feature map forms a SAR data model;
computing a Feature Domain Adaptation (FDA) loss between the plurality of electro-optical encoder features and the SAR data model;
comparing the computed FDA loss with a threshold of FDA loss; and
retraining the SAR encoder till the computed FDA loss is below a threshold, if the computed FDA loss is exceeding the threshold of FDA loss.

2. The system as claimed in claim 1, wherein each of the plurality of NeWA modules generates the weight value corresponding to the feature map, by:
obtaining a focused feature map by performing feature extraction from the feature map, at a first, a second, and a third convolution layer of the NeWA module, wherein the focused feature map is a distilled knowledge of the feature map that is input to the NeWA module; and
generating a feature vector corresponding to the focused feature map, wherein the feature vector is a 1-Dimensional weight vector corresponding to the feature map that is input to the NeWA module.

3. A processor implemented method (400) of domain adaptation, comprising:
training (402) an electro-optical encoder and a pretrained data model by performing LULC classification of a plurality of electro-optical data, to generate a LULC classification map, wherein the LULC classification map comprises a plurality of electro-optical encoder features;
training (404) an SAR encoder using SAR data, freezing weights of the electro-optical encoder and the pretrained data model, comprising:
processing a feature map of the SAR data using a plurality of Neural Weighted Averaging (NeWA) modules of the SAR encoder, wherein each of the plurality of NeWA modules generates a weight value corresponding to the feature map;
generating a consolidated feature map at each of the plurality of NeWA modules by applying a plurality of the weight value generated by the plurality of NeWA modules on the feature map of the SAR data; and
generating an output feature map by concatenating a plurality of the consolidated feature maps weight value generated by the plurality of NeWA modules, at each of the plurality of NeWA modules, wherein,
the output feature map forms a SAR data model;
computing (406) a Feature Domain Adaptation (FDA) loss between the plurality of electro-optical encoder features and the SAR data model;
comparing (408) the computed FDA loss with a threshold of FDA loss; and
retraining (410) the SAR encoder till the computed FDA loss is below a threshold, if the computed FDA loss is exceeding the threshold of FDA loss.

4. The method as claimed in claim 3, wherein generating the weight value corresponding to the feature map by each of the plurality of NeWA modules comprises:
obtaining (502) a focused feature map by performing feature extraction from the feature map, at a first, a second, and a third convolution layer of the NeWA module, wherein the focused feature map is a distilled knowledge of the feature map that is input to the NeWA module; and
generating (504) a feature vector corresponding to the focused feature map, wherein the feature vector is a 1-Dimensional weight vector corresponding to the feature map that is input to the NeWA module.

Dated this 19th Day of July 2022

Tata Consultancy Services Limited
By their Agent & Attorney

(Adheesh Nargolkar)
of Khaitan & Co
Reg No IN-PA-1086