Research / Internships / Projects
How Visual Cues Fuel Emotional States, Engagement and Retention in Children’s Educational Videos
Tara Gangwar, Prof. Sumeet Kumar, Indian School of Business
ISB Research Synopsis
The research study involved studying more than 30,000 videos from the top 100 YouTube children’s channels and found that videos with more visual motion tended to get higher
I analysed how children engage with video content using eye-tracking technology. Reviewed sample video recordings of the 53 participating children and coded the emotions expressed while they engaged with the videos. Employed the DeepFace library to identify and classify children’s emotional expressions. Explored how the facial emotion captured while viewing the video correlated to the emotional state expressed by the children soon after viewing each video.
- Analysed how children engage with video content using eye-tracking technology.
- Reviewed sample video recordings of the 53 participating children and coded the emotions expressed while they engaged with the videos.
- Employed the Open CV, DeepFace library to identify and classify children’s emotional expressions.
- Explored how the facial emotion captured while viewing the video correlated to the emotional state expressed by the children soon after viewing each video.
- Its an ongoing research work, findings of the research will be published in peer-reviewed journal soon.
Facial Emotion Recognition and Gesture Movements in Autistic Kids
AiSPRY Internship
1. AI Platform to Detect Facial Emotion Recognition and Gesture Movements in Autistic Kids
- Used real-time facial emotion recognition software which was used to build an educational platform for autistic and neurodivergent children.
- Reviewed existing literature on ASD, correlating to eye movements and gestures with autism and identifying facial structures most common among autistic children.
- Created a dataset from open-source videos and developed an algorithm using Roboflow to categorize varied eye movements and gestures of ASD and normal children.
- This work enabled differentiation between typical and autistic patterns. This work has been adopted by AISPRY for creating effective educational tools for autistic children.
2. Saathi LLM Chatbot- Empowering Women to Gain Access Financial Information
- In addition, built a LLM Chatbot- ‘Saathi’ for helping rural women to anonymously get access to information- female health, financial data, etc.
- In future, plan to make a conversational based as well as vernacular chatbot
Facial Emotion Recognition
INSPIRIT AI Summer Fellow
Built a AI based Facial emotion detection using machine learning algorithms and deep learning. Used publicly available datasets and OpenCV and Deepface in Python.
- Made a Facial Emotion Detection AI platform using machine learning algorithms and deep learning models- loading data, building and testing model
- Used publicly available dataset to train the model to detect various emotional states- happy, sad, angry
- Analysed emotions from faces in frame, regions of interest
- Used Open CV, Deepface in Python
Cost-effective Vape Detector
Built & Tested a Cost-Effective Vape Detector.
- Used Raspberry Pi (model 3,4), gas sensors and humidity sensors (to detect smoke), analog to digital converter, programmed a LED alerting system.
- Beta prototype was setup in school bathroom to test possible vaping amongst students.
Saathi LLM Chatbot
Saathi LLM Chatbot- Empowering Women to Gain Access Financial Information
- In addition, built a LLM Chatbot- ‘Saathi’ for helping rural women to anonymously get access to information- female health, financial data, etc.
- In future, plan to make a conversational based as well as vernacular chatbot
Correlation of Earthquakes Occurrence based on Phases of the Moon
Astrological Influences On Earthquakes: Myth Or Reality?
Research Associate under Prof. Vikram Pudi, International Indian Institute of Technology (IIIT Hyderabad)
Published a paper on correlation between Earthquakes and Astrological events. Used Python and libraries like -Pandas, NumPy, and Sklearn for research.
