We present the “AI-Atlas” didactic concept as a coherent set of best practices for teaching Artificial Intelligence (AI) and Machine Learning (ML) to a technical audience in tertiary education, and report on its implementation and evaluation within a design-based research framework and two actual courses: an introduction to AI within the final year of an undergraduate computer science program, as well as an introduction to ML within an interdisciplinary graduate program in engineering. The concept was developed in reaction to the recent AI surge and corresponding demand for foundational teaching on the subject to a broad and diverse audience, with on-site teaching of small classes in mind and designed to build on the specific strengths in motivational public speaking of the lecturers. The research question and focus of our evaluation is to what extent the concept serves this purpose, specifically taking into account the necessary but unforeseen transfer to ongoing hybrid and fully online teaching since March 2020 due to the COVID-19 pandemic. Our contribution is two-fold: besides (i) presenting a general didactic concept for tertiary engineering education in AI and ML, ready for adoption, we (ii) draw conclusions from the comparison of qualitative student evaluations (n = 24–30) and quantitative exam results (n = 62–113) of two full semesters under pandemic conditions with the result of previous years (participants from Zurich, Switzerland). This yields specific recommendations for the adoption of any technical curriculum under flexible teaching conditions—be it on-site, hybrid, or online.
12 December 2024 in Computers in Biology and Medicine
Total Hip Arthroplasty (THA) is a well-established and common orthopedic surgery. Due to the complexity involved in THA, orthopedic surgeons require rigorous training. However, the current gold standard, the tutor-guided and -evaluated apprenticeship model is time-consuming, costly, and poses risks to patients. There is a pressing need for additional training resources to enhance the efficiency and safety of the training process. In this work, we present a novel Augmented Reality (AR)-based simulator designed for THA that helps enable a new self-paced training and learning paradigm without the need for instructors. The simulator reduces the need for instructors by integrating an AR guidance module and an automated performance evaluation module. Three types of AR guidance were developed: Overlay, Virtual Twin, and Sectional Views. A feasibility study was conducted with five resident surgeons and two senior surgeons to compare these guidance methods quantitatively and qualitatively. The automated performance evaluation module was assessed against manual performance evaluation using Bland-Altman analysis with limits of agreement (LoA) and Mann-Whitney U tests. The quantitative feasibility results indicate the efficacy of the developed AR guidance, characterized by mean transitional and rotational deviation errors below 3 mm and 3 degrees. Based on the qualitative results, we provide recommendations for efficient AR guidance designs. The Bland-Altman analysis results (0.22±1.32mm with LoA -2.37 to 2.81 mm for distance deviation, 0.94±2.41 degrees with LoA -3.78 to 5.66 degrees for yaw deviation, -0.34±1.30 degrees with LoA -2.90 to 2.22 degrees for pitch deviation) and p-values of Mann-Whitney U tests (0.64 for distance deviation, 0.12 for yaw deviation, 0.11 for pitch deviation) indicate no statistically significant differences between the automated and manual performance evaluation at a significance level of 0.05. This work shows the potential of AR-based simulators in introducing a novel, data-driven approach to open surgery training in orthopedics, enabling surgeons to individually assess and improve their progress.
Automatic detection of persons is an important application in visual surveillance. In general, state-of-the-art systems have two main disadvantages: First, usually a general detector has to be learned that is applicable to a wide range of scenes. Thus, the training is time-consuming and requires a huge amount of labeled data. Second, the data is usually processed centralized, which leads to a huge network traffic. Thus, the goal of this paper is to overcome these problems, which is realized by a person detection system, that is based on distributed smart cameras (DSCs). Assuming that we have a large number of cameras with partly overlapping views, the main idea is to reduce the model complexity of the detector by training a specific detector for each camera. These detectors are initialized by a pre-trained classifier, that is then adapted for a specific camera by co-training. In particular, for co-training we apply an on-line learning method (i.e., boosting for feature selection), where the information exchange is realized via mapping the overlapping views onto each other by using a homography. Thus, we have a compact scene-dependent representation, which allows to train and to evaluate the classifiers on an embedded device. Moreover, since the information transfer is reduced to exchanging positions the required network-traffic is minimal. The power of the approach is demonstrated in various experiments on different publicly available data sets. In fact, we show that on-line learning and applying DSCs can benefit from each other.
Integrating drones into the civil airspace is one of the biggest challenges for civil aviation, responsible authorities and involved com- panies around the world in the upcoming years. For a full integration into non-segregated airspace such a system has to provide the capability to automatically detect and avoid other airspace users. Electro-optical cameras have proven to be an adequate sensor to detect all types of aerial objects, especially for smaller ones such as gliders or paragliders. Robust detection and tracking of approaching traffic on a potential collision course is the key component for a successful avoidance maneuver. In this paper we focus on the aerial object tracking during dynamic flight maneuvers of the own-ship where accurate attitude information corresponding to the camera images is essential. Because the ’detect and avoid’ functionality typically extends existing autopilot systems the received attitude measurements have unknown delays and dynamics. We present an efficient method to calculate the angular rates from a multi camera rig which we fuse with the delayed attitude measurements. This allows for estimating accurate absolute attitude angles for every camera frame. The proposed method is further integrated into an aerial object tracking framework. A detailed evaluation of the pipeline on real collision encounter scenarios shows that the multi camera rig based attitude estimation enables the correct tracking of approaching traffic during dynamic flight, at which the tracking framework previously failed