Magnetic Reconnection and Generation of Radio waves in Star-Planet Interaction

---

Term Paper (Electrodynamics)

Term paper completed during the Electrodynamics course under the guidance of Dr. Rahul Sharma.

Facial Emotion Detection

---

End Term project for Introduction to Data Science 2

Semester Project completed during the Introduction to Data Science (IDC) 1 course under the guidance of Prof. P. Radha Krishna.

In this project, we developed a facial emotion detection model using deep neural networks, specifically convolutional neural networks (CNN), to classify human emotions from images. Facial emotion detection is a critical technology today, with applications ranging from driver's drowsiness detection to analyzing student behavior in educational environments.

Approach & Methodology:

We used Keras with TensorFlow backend for building the CNN, leveraging key libraries for image processing and network architecture. The project involved the following steps:

1. Data Preparation:
   

   We utilized ImageDataGenerator to handle both training and validation datasets. Images were resized to 48x48, converted to grayscale, and batch processing was set to 64. The dataset was categorized into seven classes of emotions.



2. CNN Architecture:
   

   • We implemented a sequential model in Keras. The model consisted of several convolutional layers, each followed by max-pooling and dropout to prevent overfitting.

   • The first convolutional layer had 32 filters with a 3x3 kernel and ReLU activation. The grayscale input was shaped as 48x48x1.

   • Subsequent layers used 64 filters with similar kernel sizes and pooling layers.

   • A fully connected dense layer with 1024 units followed the flattening of feature maps, with a dropout rate of 0.5 to reduce overfitting.

   • The final output layer used a softmax activation to classify the input into one of seven emotions.



3. Training the Model:
   

   We compiled the model using categorical_crossentropy as the loss function, with Adam as the optimizer (learning rate 0.0001, decay of 1e-6), and accuracy as the evaluation metric. The model was trained for 50 epochs, and validation was performed using the test data.



4. Saving the Model:
   

   After training, we saved the model's architecture as emotion_model.json and its learned weights in an emotion_model.h5 file.



5. Real-time Emotion Detection:
   

   For real-time emotion detection, we used OpenCV to access the webcam or video feed. The model was loaded and applied to detect faces in the video using the Haarcascade classifier. Detected faces were resized to 48x48 and passed through the CNN for emotion prediction. The predicted emotion was displayed in the top left corner of the face detection box in real-time.

Conclusion:

This project demonstrates how convolutional neural networks can effectively classify human emotions from facial expressions, with potential applications in several real-world scenarios such as surveillance, education, and automotive safety.

GitHub Repository:
You can find the project code and more details on my GitHub repository.

Description Video:

Detecting Pulsar Stars

---

End Term project for Introduction to Data Science 1

Semester Project completed during the Introduction to Data Science (IDC) 1 course under the guidance of Prof. P. Radha Krishna.

The Three Body Problem

---

End Term project for Numerical Methods and Programming course

Semester Project completed during the Numerical Methods and Programming course under the guidance of Dr. Achanta Venugopal.

Trajectory of three bodies

Trajectory with step size 0.0001

Introduction to Galaxies and its several parts

---

Talk: NAXATRA, Indian Institute of Science Edcuation and Research Berhampur Astronomy Club

During my 1st year at IISER Berhampur, I delivered a talk on the fascinating topic of galaxy formation and classification. I explored how galaxies, as homes for stars and other celestial objects, form through gravitational forces and the bottom-up theory, which suggests that star clusters merge into larger galaxies. The presentation covered various galaxy types, including elliptical, spiral, and lenticular galaxies, detailing their unique characteristics and classifications. Additionally, I discussed the concept of galaxy clusters and superclusters, along with intriguing examples like Seyfert galaxies and Hoag's Object.

Stellar Evolution: The Life Cycle of a Star

---

Talk: NAXATRA, Indian Institute of Science Edcuation and Research Berhampur Astronomy Club

During my 2nd year at IISER Berhampur, I delivered this talk on the complex stages of stellar evolution, from the formation of stars in interstellar clouds to their final stages as white or black dwarfs. I covered key phases such as protostar formation, main sequence nuclear fusion, and the transition into red giants, leading to planetary nebulae and dense remnants. The talk also highlighted unique phenomena like helium flashes, protostellar winds, and the rare event of a Nova.

Solar Model: From Conceptualization to Presentation

---

STREAM 2020: Science Outreach Program (IISER Berhampur)

During my 1st year at IISER Berhampur, I led a team presentation in a science outreach program, where we created an educational model of the Sun. In this presentation, we explored the Sun's basic structure and introduced various solar phenomena, including sunspots, spicules, coronal mass ejections (CMEs), and prominences. We discussed differential rotation and the Sun's inner layers, highlighting the convection flow and the science behind sunspots. Additionally, we covered key concepts such as the temperature of the corona and the proton-proton chain reaction that powers the Sun. This project aimed to engage and inform the audience about the complexities of our closest star and its critical role in the solar system.

My Screen (Moire Pattern)

---

Talk: 137 Inverse (IISERBPR Physics Club)

During my 1st year at IISER Berhampur, I delivered a team presentation on the Moire Effect, focusing on the cause of this fascinating visual phenomenon. We explained how the Moire pattern emerges when two or more grids or patterns overlap at slight angles, leading to interference and the creation of new patterns. The talk covered its underlying physics, practical applications, and relevance in fields like imaging and signal processing. This presentation was part of a contest called Physicist's Unseen Battle Ground (PUBG), where I participated with my classmates.

Presentation Images: