CLICK

HERE

FOR A DEMO

• Coronavirus disease 2019 (COVID-19) is an infectious disease caused by severe acute respiratory syndrome coronavirus2.
• In the current state of the world 🦠 and with many of us in lockdown, This app gives the user a visualization of currently available data all around the globe.
• The user can choose to view the entire Global data with the infected and deaths statistics.
• The user can also view a graph of per Per Country-State-County stats. (ZOOM IN TO VIEW)
🔺 Note that the active cases isnt availavble for visualizing over the chart at the time due to API constraints.

• The user can choose to view the infected stats, or the death stats or both by simply clicking on the texts shown above the graph.
• The user can also view an entire map view of the world. -> MAP VIEW • The user can view other measures to protect themselves and follow instructions incase they are sick.

CLICK

HERE

FOR A DEMO

• This is a simple app where a user can maintain a todo list for various purposes.
✔️ Shopping
✔️ House Chores
✔️ Bills
✔️ Any Unfinished Task

• The main purpose of this app is to help organize tasks, so learning to effectively use them will benefit you by improving your time management abilities and decreasing your stress levels.
🔺 Note that this app currently doesnt have a data store thus it is a one time use app for now. (Will be modified later)

• The user can do the following:
✔️ Create a ToDo list
✔️ Add new ToDos by clicking on "ADD NEW"
✔️ Delete ToDos that are no longer needed or completed
✔️ Mark ToDos that are important or completed or saved for later

• Forest fires constitute a serious problem throughout the globe. It proposes a great risk to vegetation, wildlife, livestock and humans. This project aims to wirelessly detect the presence of fires in forest areas and predict the next possible areas for the spread using learning algorithms. Further on, it also aims in the containment of the spread using viable solutions.

• Designed a system to wirelessly detect the presence of fires in forest areas using DTH11 and TPM36 sensors in C++ and simulated that spread in Unity3D using C# and on the console using Java.

• There are two algorithms chosen to predict the fire spread once we get a sensor triggered. One is Lane Detection Algorithm and the other is Boundary Detection Algorithm. (More Details in the paper). Accuracy was also calculated for the same.

• A graph was then plotted for boundary detection algorithm and lane detection algorithm based on the number of sensors that were predicted to be the next plausible places for the fire spread and on the number of sensors that caught fire. As we see in Figure 16, boundary detection algorithm gives us an accuracy of 30 – 40%, whereas, lane detection algorithm gave an average accuracy of 80%.

• Augmented a real-world scenario that saved 1/3rd of burnt lands and $10 million.

• In this project, Supervised Learning method is used to classify the images between two classes: Cataract and Normal.
• An image classification models is designed using Convolution Neural Network (CNN) and Support Vector Machine (SVM) to detects cataract in an iris.
• TensorFlow library is used to perform computations on various images captured.
• Developed an Android Application to capture images of the iris and a designed a PHP based server to handle requests.

This shows that for few epochs, the prediction accuracy isn’t that good, in fact we get an accuracy below 50%. Ass the number of epochs is increased, the more training is performed and thus the images are better classified. As you can see, we get a pretty good accuracy of 90% with 5 epochs. But again, as the epochs are increased after a certain number the model gets overfit (giving an accuracy of 100%) and slowly we see diminishing results. Overfitting happens when a model learns the detail and noise in the training data to the extent that it negatively impacts the performance of the model on new data. This means that the noise or random fluctuations in the training data is picked up and learned as concepts by the model.

The above image shows the acuracy comparison between all the algorithms

• Developed a system using Image Processing and Machine Learning algorithms in Unity3D using C# to simulate human anatomy in a virtual environment.
This approach uses the device’s inbuilt camera to capture hand gestures instead of using additional hardware devices required for VR interaction. This captured hand gestures undergoes image processing steps which comprise of pre-processing, segmentation and feature extraction to acquire finger count. The number of fingers detected are then used to perform operations on VR objects. The entire VR environment is created using the Unity3D tool. The user only needs a Smartphone along with Head Mounted Device(HMD), which makes the system user-friendly and economical, compared to other VR interactive systems.
• The purpose of designing this system is for training and education, to help trainees and students study the human anatomy. This empowers them to learn-by-doing by directly engaging with their VR objects through immersive interactive experiences. Learning transitions from traditional memorization to giving the students the ability to apply, helps them analyse and evaluate their knowledge. This minimized the risk in operations by 30% and increased efficiency by 40%.

ACHIEVEMENTS

• Published a paper and awarded with the “Best Research Paper” at the International Conference on Emerging Trends: ICETECT '17 (Topic: "Virtual Reality: A Step Towards Telepresence"). (Patent in process for the same)

• Secured 1st position at the National Level ACM Exhibition of Exclusive Research and Academic Projects held by I2IT.

• Best Project Award - Persistent Systems.