Antika Roy

🔗 Google colab notebook link
Description: I participated in 'Fatima Fellowship Quick Coding Challenge'. The task is to detect the cat or dog in the image is straight or upside down. I have used the dataset from Kaggle 'Cats vs dogs' dataset. With the ResNet18 model I got an accuracy of 99.90% in 5 epochs with three incorrect predictions.
Technology : Python, Google Colab, PyTorch

🔗 Kaggle Competetion link
🔗 Google colab notebook link
Description: In this competition, I have trained my model on a novel semantic similarity dataset to extract relevant information by matching key phrases in patent documents. Determining the semantic similarity between phrases is critically important during the patent search and examination process to determine if an invention has been described before.
Technology : Python, Kaggle Notebook, PyTorch

🔗 Kaggle Competetion link
🔗 Google colab notebook link
Description: In this competition, , I had to identify key phrases in patient notes from medical licensing exams using RoBERTa model where the system learns to predict intentionally hidden sections of text within otherwise unannotated language examples.
Technology : Python, Kaggle Notebook, PyTorch

🔗 Github link
Technology : python, east text detection, tesseract OCR, tkinter gui
This is an upgraded work of my undergrad thesis project on "Text detection from natural scene images". I have used OpenCV package which uses the EAST model for text detection and the tesseract package is for recognizing text in the bounding box detected for the text. My image set contains natural scene text detection challenges and EAST text detector can successfully detect the bounding boxes having the text.
Key Points :

• Loading Pre-trained EAST model and defining output layers and forward pass the image
• Derive the bounding boxes after applying non-max-suppression
• PSM for the Tesseract has been set accordingly to the image.

🔗 Github link
Technology : python, scikit-learn, numpy, matplotlib
Description: This is implementation of the K Means algorithm to classify hand written digits on MNIST digit dataset.MNIST digit dataset essentially contains pixel data of bunch of different images of numbers values that are handwritten from 0 to 9.K Means Clustering alrogithm classify these digits based on the pixel data.The challenge was to feed these two dimensional pixel valus to our model.
Key Points :

• Importing Modules sklearn,matplotlib,numpy
• Loading the MNIST Data-set
• Scoring and training the model
• MatplotLib Visualization

🔗 Github link
🔗 Dataset used
Technology : python, scikit-learn, svm
Description: I have trained my model with Support vector machine classifier on a breast cancer dataset which consists features describing a tumor cell nucleus with real-valued features like radius (mean of distances from center to points on the perimeter) , texture (standard deviation of gray-scale values) , perimeter , area , smoothness (local variation in radius lengths) , compactness (perimeter^2 / area - 1.0) , concavity (severity of concave portions of the contour) , concave points (number of concave portions of the contour) , symmetry.The svm classifier model predicts the value as malignant or benign.The highest accuracy score was 0.9736842105263158
Key Points :

• Testing kernel values and soft margin
• Scoring and training the model
• Comparing accuracy values to KNearestNeighbors

🔗 Github link
🔗 Dataset used
Description: Using K-nearest neighbors classifier to classify cars as safe or unsafe based on the features in the dataset.The model evaluates cars according to the following concept structure: overall price, buying price, price of the maintenance,technical characteristics, number of doors,persons capacity ,the size of luggage boot etc.
Technology : python, pandas ,scikit-learn
Key Points :

• Training a KNN Classifier
• Converting Non numeric data
• Testing the model with different neighbour values
• Matplotlib Visualization

🔗 Github link
🔗 Dataset used
Description: Implementing the linear regression algorithm to predict students final grade based on a series of attributes like previous grades, study time, failures, absences, travel time to school etc. from student performance dataset.
Technology : python, pandas ,scikit-learn, pickle, Matplotlib
Key Points :

• Importing modules sklearn, numpy, pandas, matplotlib
• Saving Model with best score to a new file using pickle.dump().
• Training model multiple times for the best score
• Plotting final grade with features to compare by matplotlib

🔗 GitLab Repository link
🔗 Website Link
Description: Doctor management system.An online system for doctors as users and patients as third party users, creating a platform to serve and take appointment Technology : PHP, JavaScript, MySQL, Bootstrap, GitLab
Features:

• Doctors can set appointment days
• Days to date conversion
• Doctors can see patient list across a specific date
• Fake doctor removal by admin
• Online appointment system for patients