New Medium article; Query-Based Pairwise Human-Object Interaction Detection with Transformers (QPIC)(Explained)
Imagine a bustling city street captured by a surveillance camera: pedestrians crossing paths, vehicles maneuvering through traffic, cyclists weaving between lanes, and street vendors interacting with customers. For a computer vision system to make sense of this scene, it needs to not only detect the humans and objects present but also understand their interactions. This complex task is known as Human-Object Interaction (HOI) detection, a critical component for applications like autonomous driving, robotic assistance, and advanced surveillance systems.
Nov 8, 2024
Computer Vision for Detecting Potholes in Asphalt Roads in Real-Time using RC Car
Developed and trained a Convolutional Neural Network (CNN) based on the AlexNet architecture for real-time detection of potholes in asphalt roads, achieving 92.15% accuracy, 91.38% sensitivity, and an F-score of 96.52%. Optimized the model using high-performance GPUs for deployment in real-time systems.
May 1, 2024
Integrated CNN and Waterwheel Plant Algorithm for Enhanced Traffic Detection
Led a team in the design and implementation of a hybrid CNN and Waterwheel Plant Algorithm (WWPA) to detect and track vehicles in global traffic data, achieving a model accuracy of 97.28%.'
Jan 1, 2024
Face Mask Detection Using MobileNetV2
Built a deep learning model using MobileNetV2 to detect face masks, a crucial tool for public safety during the COVID-19 pandemic. Achieved a 97.71% accuracy on the test dataset, demonstrating the model’s robustness and efficiency under varying conditions.
Mar 15, 2023