Yes, we can begin.

Please start by introducing your professional journey in academia

A very good evening. I'm Dr. S Manju. And I've got nearly. 60 years of teaching experience. I started by teaching career in the year. Those in 10.

Okay Dr.

And in the college named as Velamal Institute of Technology. And then I would start. Yup. Yeah. Yes. Yeah, 2010 till 2020 I was working in Velamal Institute of Technology. And then from 2020 to 2022. I was working for. In Savita Engineering College and 2022. Starting from there till date. Working in Velammal engineering college. I'm taking handling subjects like digital signal processing and digital image processing. Where more like you want me to say the timeline in detail more?

And in the college named as Velamal Institute of Technology. And then I would start. Yup. Yeah. Yes. Yeah, 2010 till 2020 I was working in Velamal Institute of Technology. And then from 2020 to 2022. I was working for. In Savita Engineering College and 2022. Starting from there till date. Working in Velammal engineering college. I'm taking handling subjects like digital signal processing and digital image processing.

So you've spent time teaching digital signal processing and image processing. Out of all the foundational concepts you cover in those courses, which one do you enjoy teaching the most, and why?

Right, so you handled subjects like digital signal processing and image processing in a couple of different engineering colleges. Out of all the foundational topics you teach in digital image processing, is there one concept you find especially rewarding to explain to students?

Yeah, I. Can explain one topic which is. Spectral image processing.

Spectral image processing is a fascinating area. When you introduce that topic to students for the first time, what's the core idea or analogy you use to help them grasp why spectral analysis matters in image processing?

The the. Is hyperctal image processing is that? Most of the banks. Yeah.

You were talking about hyperspectral image processing and mentioned banks—can you say more about how you introduce that concept to students, or what real-world example you use to help them connect with it?

Yeah, first I used to teach them about the basics.

Fundamental of image processing. That is the difference between. A normal image processing starting from the. Central bank and then comes to the hyperspectral ban. Why? Using hyperspectral image processing is that we?

Dr. Manju, I see you've published "Optimizing Hyperspectral Image Classification Through Advanced Deep Learning Models Enhanced by Adam" in Traitement du Signal this year. Can you share how your research in hyperspectral image classification has influenced the way you design lab assignments or projects for your digital image processing students?

Given your experience in hyperspectral image processing, where do you see the most fundable research directions over the next three years? Are there particular grants or funding sources you'd target to support your work?

The main focus in this hyperspectral image processing in the funding area is to buy a hyperspectral sensor camera. Where the hypothetical sensor camera. Nearly 20 to 30,00,000. And it is really difficult for an Indian community people to get in a particular colleges to buy this. Control camera sensors and students cannot be. But many projects and all, so we have to go.

International projects where we can buy this hyperspectral sensor and we. A different pictures and then. Classification using atom optimizer and we can do even the segmentation part also so that the crop.

Right, that's a real constraint in the Indian research environment. Shifting gears—when you teach a large undergraduate image processing class, suppose you can't use slides or traditional lectures. How would you design a session that gets students actively engaged with the material?

Given your experience guiding students and managing research constraints, let's talk about institutional responsibilities. Accreditation cycles can be demanding—if you joined our department and found that outcome assessment data was inconsistent across courses, how would you approach fixing that?

Coming to the aggregation part. Worked in national board of aggregations in this NBA in criteria 2 and criteria 3 the way we. Focus more on designing and framing the course of. Come particular subject. So we used to go. Attainment and we'll have a. Discussion with the course in charges that what is a. Reason for students are not able to achieve their attainment and from that. Attainment. How can we improve?

Okay, so you've actually worked with the NBA process, especially on criteria two and three,

And so that we'll get the. Department criteria. And the outcome can be met. We will select the students as a slow learner students at the Topper students and the reason why the students are not able to. Propose outcome. So we will analyze the students learning level, understanding level in each and every courses in which. Yeah, framing. So after. Students level we. Syllabus that we can give more importance to to. Technical or. Improve practical knowledge so that we can achieve. That course outcome Ambulance.

Bus framing also. So that we can achieve the overall achievement in our department. So. See which I?

Alright,

You described analyzing students' learning levels and framing the syllabus to boost practical knowledge, which helps with course outcome attainment. But in the case where a student accuses you of grading bias and your department head is pressuring you to pass more students, what's your step-by-step approach to resolving that conflict while staying true to your academic standards?

As my ambition is to teach. Students in a good way, I will not definitely show a biasing to student which which will stop their life even in they're not able to complete properly so. But fix first with help of. We will target that is what are the ways in which a students can easily. And even if they get low marks also, we will make them to try to understand. That. We have to.

That we will not definitely fail almost with the students in subjects. So the what we used to follow.

Alright, Professor, shifting focus now—given your expertise in hyperspectral image processing,

So you’re saying you avoid failing students in your subjects and follow a particular process. Can you clarify what specific steps or guidelines you use to make sure grading is consistent and fair for everyone?

In the beginning of my course, while I'm taking class, I used to say to my students that these are the topics in which. Focus more so that the students from the beginning itself.

Particular topics and we'll be giving a small, small on the spot topics in the class itself and on the spot grading and all those things are used to practice to make students so that they won't. Difficult when they come for. Since students community, they will feel difficult to understand how to convert a paper, how to convert an idea into a paper.

You talked about structuring your first session and setting expectations. Shifting to your research guidance—can you describe a time when you helped a student take their project from an initial idea to a publishable research outcome? What was your approach at each stage?

You have to give an introduction about. Of each and every topic. And what are the drawbacks?

Okay, so you use on-the-spot topics and grading in class to help students bridge the gap from idea to paper, especially since they often struggle to turn an idea into a publication.

You emphasized the importance of introducing each topic and discussing drawbacks so students can better understand the transition from idea to paper. Can you give me an example of how you'd break down a complex research topic, like spectral unmixing in hyperspectral imaging, for students who are just starting out?

Definitely.

Definitely. So. One of an example in which one of my students are unable to identify the hyperspectral image database.

Alright, so you'd always start with a clear introduction and highlight the drawbacks, to lay the groundwork for student understanding. Now, turning to your experience with embedded and communication systems—can you share a time when you supervised a student project that involved real-time data acquisition from embedded hardware, and how you helped them troubleshoot communication issues between the sensor and processing unit?

Here the hyperspectral sensor cameras are too much costly and we cannot.

You mentioned your student struggled to identify a hyperspectral image database. How did you help them bridge that gap and get started with their research—did you guide them on sourcing data, or was there a process you followed to make it manageable?

Right, you're pointing out the high cost of hyperspectral sensor cameras. Given that limitation, how did you help your students simulate or work around the lack of direct hardware access in their embedded projects?

Sure, to begin with I use. To ask my students. To the images from the Kaggle datasets. Or from AB IRS sensors where the Indian pine datasets and Rosinas datasets are available from a research students to start with so once they down. Those data sets from the available portal research portals so they can. The preprocessing stage and then. Classification and followed by the segmentation. So beginning since we don't have a cap. Sensors we used to collect the data from the data set which is available in the web so that my students can. Informations. And they have to. High dimensional reduction. Problem which is existing in this high. Images.

Okay, so you guide your students to source datasets from platforms like Kaggle or research portals, and then move through preprocessing, classification, segmentation, and dimensionality reduction challenges. Shifting gears a bit, can you tell me how you structure your lectures or labs to ensure your delivery is clear and students actually grasp complex technical concepts, especially when they're new to the material?

We have analyzed first what is a type. Features should be. From that. Trouble.

So you start by analyzing what features to extract from the images and then move from there. Can you give me a concrete example of how you break down a complex process—like feature extraction in high-dimensional hyperspectral data—so that students can follow each step clearly in a lab or lecture?

I used to follow a PowerPoint presentation first in the beginning of my class to distinguish a multispectral and a hyperspectral images and then I will give. Nude images or five to 10 images datasets and from that data. They have to identify the type.

The weight of filters are more and more like that so that the students get.

Alright, so you use PowerPoint to clarify multispectral versus hyperspectral images, then have students work hands-on with a small dataset to identify types. When you're conducting exams or evaluating students, how do you ensure the process is objective and aligns with university policies?

You were saying the weights of filters are more, and you were about to explain how that affects your students. Go on.

Yeah, that is the filters. What I the Veena filters, it is not weight, it's veiner filter which is used for pre. State because in hyper. Since it deals with more than thousands of images.

Filtering cannot die. So the term where are. The preprocess.

Alright, so you use Wiener filters for preprocessing because hyperspectral images involve thousands of bands and noise is a real issue. When you explain this to students,

Okay, so you’re saying filtering is an essential part of preprocessing and can’t be skipped, especially given the complexity of hyperspectral data. How do you help students actually see the impact of a filter like the Wiener filter—do you have them run experiments or compare results visually?

Yes, I have one of my student who worked in Indian pine data set. Or Indian Salinas data set which is one of the best hype? Spectral data set example and when they worked in Indian pine data set they have down. That image and they have worked. In one of the. Filtering Vena filtering to preprocess.

So your student worked with the Indian Pines or Salinas hyperspectral datasets and used Wiener filtering for preprocessing.

And after preprocessing they extract the spatial and. Features from. Bye bye. From the. So it is papers like the best. Classifiers. It starts from. Victor Mission SVM classifiers. The beginning in SVM and then it goes to the. Clevance Victor Mission and then comes to the deep learning or mission learning algorithm. So you go.

And after preprocessing they extract the spatial and. Features from. Bye bye. From the. So it is papers like the best. Classifiers. It starts from. Victor Mission SVM classifiers. The beginning in SVM and then it goes to the. Clevance Victor Mission and then comes to the deep learning or mission learning algorithm. So you go. Literary way we'll go for. A comparison between SVM then RVM or an e-mail. Relevant Specta mission and then we'll choose a best class. Fire and after. Notification. An album like a particle swamp. Optimization algorithm or the best? Of algorithm to opt. Clarified image. And then? For the segmentation process. It's only.

You mentioned guiding a student through preprocessing using Wiener filtering on the Indian Pines dataset, but I'd like a bit more detail. Can you walk me through how you actually supervised their research process—like helping them frame a research question, structure their experiments, or troubleshoot technical roadblocks along the way?

We are able to understand that. RPM.

You walked through how your students compare SVM, RVM,

Okay, so you mentioned RVM—relevance vector machine—and that your students are able to understand its application. When you're evaluating their performance on exam questions tied to these advanced classifiers, how do you ensure fair and consistent grading across students?

We will conduct a Viva question individually. We and we used to frame a rubrics based upon the team. Performance and the individual performance so individually we will ask the students about the. Dear View and the Knowledge. They gathered through the literature, so one. Individual performance is calculated then as a. When they're submitting a project, a group either with a three batch or three or four together forming A1 batch.

We will conduct a Viva question individually. We and we used to frame a rubrics based upon the team. Performance and the individual performance so individually we will ask the students about the. Dear View and the Knowledge. They gathered through the literature, so one. Individual performance is calculated then as a. When they're submitting a project, a group either with a three batch or three or four together forming A1 batch. It based upon the team performance.

That individ. So you're getting the information from the students and then?

Alright, we're wrapping up. Is there anything you'd like to ask about the role or the team?

Right, so you run individual viva sessions and use rubrics to score both individual and team project performance, which keeps things structured and fair. Alright, we're wrapping up. Is there anything you'd like to ask about the role or the team?

Right, you’re asking if we gather information from students individually during assessment. Yes, that’s exactly the idea—through viva sessions and rubrics, we make sure we’re evaluating what each student truly understands, not just what the team submits. Now, is there anything you’d like to ask about the role itself or the team you’d be joining?

I would like. Like to know that the? Role which I'm going to join is about the assistant professor.

Yes.

Of course, go ahead with your question.

Got it.

Yeah. I won't.

Alright.

Yeah, I would like to join as an assistant.

You'd like to join as an assistant professor—thanks for sharing that. Before we close, is there anything else you'd like to know about the department or the expectations for this role? If not, I'll wrap up the conversation.

Just give me what are the? Knowledge I have to develop so that it'll be easy for. No, I just would like to know about what. Extra knowledge I have to gather so that it will be easy for me to develop. My knowledge, skills into come up to the department.