Insights, tutorials, and deep dives into computer vision
Facial Keypoint Detection is a crucial task for facial recognition software applications.
In this note, I explore the use of Convolutional Neural Networks (CNNs) to develop a classification model capable of predicting COVID-19 cases based on x-ray images.
This note develops an End-2-End pipeline and deep learning model to predict the presence of Pneumonia on chest X-ray images.
When building computer vision models, a common task in the data processing pipeline involves performing data transformation and augmentation using transforms. This is a demonstration for their implementation.
Social Media apps like Instagram and Snapchat offer a number of filters to create realistic effects on video and image content. Here is a pencil sketch implementation.
This note demonstrates how to use out-of-the-box Haar Cascades model for Face Detection on images.
R offers a number of packages that build on top of ggplot2 to generate themes and visualize that can enhance storytelling and data presentation.
Sobel filters are a powerful way to amplify the vertical and horizontal boundaries on an image
Bayesian forecasting methods are an effective alternative to traditional frequentist forecasting methods, such as ARIMA methods, as they offer more transparent parameter estimation and robust models.
Masking is a technique used in computer vision that can achieve results similar to chroma keying also known as green screen background replacement in video and filmmaking.
Histogram equalization uses all pixel values to improve the contrast of an image. While it can improve contrast through intensity distribution, it can sometimes fail, leading to overly bright images.
Histogram Equalization is the process of normalizing the intensity of the individual pixel against its respective probability (the likelihood of that pixel intensity value in the image).
In the previous note, we discussed the theoretical implementations of Kernel Convolutions. This short note demonstrates kernel convolutions with OpenCV.
The foundation concept for processing, manipulating, and transforming images is a Kernel and the use of kernels in the convolution process.
This short note breaks down the Binary Cross Entropy Loss function often used in classification tasks in deep learning.
This note explores Logistic Regression implementation in Pytorch.
This note explores the Pytorch Library using a simple linear regression case.
Generating datasets for training and testing machine learning algorithms can be an effective way to broaden your understanding of the nuances of different ML techniques.
In this final notebook, we cover the implementation of advanced techniques for sentiment analysis using modern techniques in deep learning using Dense Neural Networks and Recurrent Neural Networks.
This note goes beyond traditional techniques for numerical text representation to Distributed Representation.
On this note, we move on to building Sentiment Classifiers using Naive Bayes and Support Vector Machines - Linear Classifiers.
This note builds on Part 1, to discuss feature extraction techniques for Natural Language Processing. Broadly, this section deals with moving from text to numerical representation.
This 5-part series covers concepts, analysis, and Machine Learning Models for Natural Language Processing.
An implementation of the K-NN algorithm for classification in Python
This note explores common non-linear functions used for solving diminishing returns optimization problems.
Working as a data scientist in digital marketing, a significant amount of my time is spent on building cost-to-revenue optimization. Example implementation in Python.
While closed-form solutions for computing coefficients are neat, inverting a matrix is computationally prohibitive, therefore, gradient descent is a computationally feasible implementation.
A foundation introduction and implementation of the Autoregressive Process and AR(p) model with Python.
Implementation of the Foundations of Moving Average Model for Time Series Forecasting in Python