Computer vision as a subfield of AI has been around for several years dealing with how computers can understand from digital images and video sequences. Advanced computer vision algorithms have already demonstrated successful applications in a variety of domains, including medical image interpretation, remote surgery, surveillance systems, security and biometrics, autonomous vehicles, and scene reconstruction, purposing to name a few. There is a list of fascinating problems in applied computer vision in medical imaging, with object detection and localization being one of the most interesting ones. Object detection and localization is now also widely associated with self-driving cars where automatic systems combine computer vision, LIDAR, and GPUs to generate a multidimensional representation of the road with all its participants. It is also commonly used in medical image analysis, video surveillance and monitoring, counting people for general statistics, and computationally analyze customer experience with walking patterns within shopping centers. In this summer school, you will learn -from scratch- how to use advanced computer vision algorithms to tackle the problem of object detection and localization in medical images. We will discuss object detection mechanism(s) in practice with several hands-on-practices starting from manual image annotation to programming and implementation in Python. We, together, will explore what object detection computational vision algorithm is, what is does, and how. The current mini summer camp at the University of Pittsburgh is structured such that in addition to attending lectures, the students will be also working in teams on a project assignment. Topics included but not limited to: - Introduction to Computer Vision - Introduction to Deep Learning Computer Vision - Deep Convolutional Neural Networks (CNNs) - Introduction to Object Detection and Localization in Computer Vision - Introduction to PyTorch - Manual Annotation of Medical Images using the LabelImg Toolset - Sliding Windows and Bounding Boxes in Object Detection - Non-max Suppression - YOLO (You Only Look Once) and SSD (Single Shot Detector) 25 seats in-person and 25 seats virtual (Zoom) are available on a first-come-first-serve basis. Co-sponsored by: Dr Ahmad Tafti, Pitt HexAI Research Laboratory at the University of Pittsburgh School of Health and Rehabilitation Sciences Speaker(s): Ahmad, Soheyla Agenda: Schedule: Day Time Agendas Monday, July 10th 9:30 – 10:30 Introduction to Computer Vision 10:30 – 10:45 Break 10:45 – 11:45 Introduction to Deep Learning 11:45 – 12:00 Break 12:00 – 13:00 Hands-on-Practice: Google Colab; What and Why? Tuesday, July 11th 9:30 – 10:30 Deep Convolutional Neural Networks (CNNs) 10:30 – 10:45 Break 10:45 – 11:45 Introduction to PyTorch 11:45 – 12:00 Break 12:00 – 13:00 Hands-on-Practice: Medical image annotation (manual annotation) using LabelImg Wednesday, July 12th 9:30 – 10:30 Sliding Windows and Convolutional Implementation of Sliding Windows 10:30 – 10:45 Break 10:45 – 11:45 Bounding Box Prediction and Intersection Over Union (IoU) 11:45 – 12:00 Break 12:00 – 13:00 Hands-on-Practice: OAI Imaging Dataset (https://nda.nih.gov/oai) plus Pizza and soft drinks!!! Thursday, July 13th 9:30 – 10:30 Non-Max Suppression, YOLO (You Only Look Once) and SSD (Single Shot Detector) 10:30 – 10:45 Break 10:45 – 11:45 Hands-on-Practice: Detection and localization of Total Knee Arthroplasty (TKA) implants in plain X-ray images 11:45 – 12:00 Break 12:00 – 12:30 Hands-on-Practice: Model analysis; IoU Friday, July 14th 9:30 – 10:30 · Project Definition and Team Building · Teams will start working on their projects 10:30 – 10:45 Break 10:45 – 11:45 Teams will be working and finalizing their projects 11:45 – 12:00 Break 12:00 – 13:00 Project Presentation and Pizza!!! Room: 6048, Bldg: Forbes Tower, 219 Meyran Avenue, Pittsburgh, Pennsylvania, United States, 15213, Virtual: https://events.vtools.ieee.org/m/362022
