Collegiate basketball is characterized by high-impact movements such as jump landings, making athletes more susceptible to injuries. Critical biomechanical factors like knee flexion, lateral trunk flexion, and foot landing asymmetry are strongly associated with injury risk. This study aims to predict six biomechanical risk factors in the landing error scoring system (LESS). The dataset comprises 8600 video frames of counter-movement jumps (CMJs) from 17 NCAA Division I female basketball athletes, recorded from frontal and lateral perspectives and annotated using a customized error annotation algorithm. The study uses the You Only Look Once (YOLOv5nu) model to analyze the basketball athletes’ CMJ videos. It demonstrates high reliability in predicting risk factors with an average Box Precision (Box P) of 0.800, recall (R) of 0.877, and mean Average Precision at IoU threshold 0.5 ([email protected]) of 0.879.

The increase in the use of autonomous vehicles has motivated a paradigm shift in the transportation domain as it redefines the boundaries of urban mobility by augmenting safety measures. This research paper explores an innovative approach to enhance 6G millimeter-wave (mmWave) beam prediction for vehicle-to-infrastructure (V2I) communications by using generative adversarial networks (GANs). By generating synthetic data samples effectively and balancing the real-world dataset, we improve the accuracy of beam prediction models significantly. Our proposed method of training random forests on synthetic data (RFGAN) to predict beam indices provides the solution for imbalanced class issues and significantly improves the predictive performance of mmWave beam selection, contributing to more reliable and efficient V2I communications. This work also performs comparative analysis with state-of-the-art models in top-K evaluation metrics, average power loss, and overhead savings related to adapting to the new approach.

Atmospheric corrections (AC) factors in adverse atmospheric effects while determining surface reflectance (SR) from remote sensed satellite images. Deep learning (DL)-based methods using only remote-sensed images are capable of learning nonlinear relations between the top of the atmosphere (TOA) and the bottom of the atmosphere (BOA) to provide consistent SR estimates. The changes over time in atmospheric conditions, geographical terrains and sensing processes has caused significant changes in the statistical properties of satellite images affecting the nature of the relationship between TOA and BOA, commonly referred to as a dataset shift. If this phenomenon is significant then over time the statistical properties of the data used to train DL models are vastly different from the test data for which we are performing AC. This distributional mismatch can cause DL models to perform poorly due to its inability to extend temporally. In this paper, we aim to ascertain the presence of data set shift using Landsat 8 images from 2013 and 2020. We create a joint distribution between the TOA and BOA for temporally separated training and testing sets, and study data set shift from three perspectives; 1. Use the Kolmogorov-Smirnov test to measure the distribution changes. 2. Apply Wasserstein distance over differences in TOA and BOA to measure the numerical drift directly. 3. Using the DL model to identify the drift and study model generalization. All three approaches indicate a dataset shift and provide strong evidence for its presence.

Using Data Analytics is a vital part of sport performance enhancement. We collect data from the Division 1 'Women's basketball athletes and coaches at our university, for use in analysis and prediction. Several data sources are used daily and weekly: WHOOP straps, weekly surveys, polar straps, jump analysis, and training session information. In this paper, we present an online dashboard to visually present the data to the athletes and coaches. R shiny was used to develop the platform, with the data stored on the cloud for instant updates of the dashboard as the data becomes available. The performance of athletes can be compared to the group averages, while coaches have access to all athletes and can compare them to each other and the team averages for all parameters. A simple color-coded design was utilized to convey the coaches which of the measured parameters is in an acceptable range and which is deficient. The dashboard was reviewed by the athletes, coaches, and exercise scientists and was useful for their needs.

Competitive sports require rapid and intense movements, such as jump landings, making athletes susceptible to injuries due to altered neuromuscular control and joint mechanics. Biomechanical features during landings are associated with injury risk, emphasizing proper movement and postural stability. Computer vision techniques offer a time-efficient, noninvasive, and unbiased method to assess jump-landings and identify injury risks. This study proposes a video analysis framework to evaluate jump landing biomechanics in athletes to determine irregular movements and incorrect postures. It provides advice and recommendations to coaches for injury prediction and training improvements. The proposed framework is tested using countermovement jump videos of 17 NCAA Division I female basketball athletes. The results indicated a low Mean Absolute Error (0.97), high correlation (0.89), high average accuracy (98.31%) and F1 score (0.98), signifying the framework’s reliability in identifying injury risk.

Person retrieval video using natural language description (NLD) is an emerging research area and depends largely on dataset diversity. Unifying datasets increases overall quality; therefore, the paper presents a case study on merging two different style data sets; one has NLD with images (CUHK-PEDES), and the other has discrete annotations with videos (AVSS). The unifying framework brings out the practical challenges and their solution. Explicit discussions on data set merging frameworks are missing in the literature, and our work will facilitate the researchers’ requirements.

Atmospheric correction (AC) is the process of retrieving correct surface reflectance (SR) values from the top of the atmosphere (TOA) radiance values by removing the effect of the atmosphere from the electromagnetic radiation reflected from the earth. Many physics-based approaches perform AC. However, these approaches, due to the complex relationship among the factors involved in AC, are compute intensive and rely on precomputed lookup tables that are based on approximations of the true relationships among these factors. The rapid growth in computational power, advancements in remote sensing technology, availability of vast amounts of satellite imaging data, coupled with advances in tools, techniques and algorithms in machine and statistical learning, has resulted in an opportunity to employ Deep Learning (DL) based approaches for providing effective solutions for AC. In this paper, we explore the potential of deep learning for AC. We categorize and review three approaches for AC: DL assisted physics-based approach, physics aware DL approach and physics agnostic DL approach. The paper overviews each of these approaches by providing the rationale behind them, and discusses key open issues and highlights possible solutions in different contexts for AC using DL.

Researchers are increasingly interested to study novel attacks on machine learning models. The classifiers are fooled by making small perturbation to the input or by learning patches that can be applied to objects. In this paper we present an iterative approach to generate a patch that when digitally placed on the face can successfully fool the facial recognition system. We focus on dodging attack where a target face is misidentified as any other face. The proof of concept is show-cased using FGSM and FaceNet face recognition system under the white-box attack. The framework is generic and it can be extended to other noise model and recognition system. It has been evaluated for different - patch size, noise strength, patch location, number of patches and dataset. The experiments shows that the proposed approach can significantly lower the recognition accuracy. Compared to state of the art digital-world attacks, the proposed approach is simpler and can generate inconspicuous natural looking patch with comparable fool rate and smallest patch size.

Physical characteristics or soft biometrics are visually perceptible aspects of a human body. Noticeable attributes like build, height, complexion, clothes help with the development of a human surveillance system. The paper proposes Discrete Soft biometric Attribute-based Person Retrieval (DSA-PR) from a video using height, gender, torso (clothes) color-1, torso color-2, and torso (clothes) type given in a textual query. The DSA-PR uses Mask R-CNN for semantic segmentation and ResNet-50 for attribute classification. Height is estimated using the Tsai camera calibration method. DSA-PR weighs attributes and fuses their probability to generate a final score for each detected person. The proposed approach achieves an average Intersection-over-Union (IoU) of 0.602 and retrieval with IoU $\ge$ 0.4 is 0.808 over the AVSS challenge II dataset which works out to 5.8% and 2.02% above the state-of-the-art techniques respectively.

Active learning is an effective method for domain knowledge and life skills learning. However, it demands more resources compared to classical teaching oriented pedagogical approaches. The paper proposes implementation of active learning approaches with the help of a real-world project, spanning across multiple courses, offered by different faculty. The paper describes the goals, design constraints, resource utilization, projects content, deliverable, implementation methodology, and quantitative analytics. Further, the paper presents analysis of over three years of observations, carried out with more than 200 student participants. Statistical analysis and student satisfaction surveys reveal that the proposed project across courses is able to increases the intrinsic motivation in students while successfully retaining the concepts delivery under constrained environments.

This paper studies and quantifies the impact of active learning experienced through multinational projects. The hypothesis was engineering education delivered through Active Learning in multicultural environment improves student competencies. The investigation captures the impact of international exposure program in developing global competencies of the modern engineer. The paper shows positive trends in the development of domain and life skills of engineering students. Post-survey after six months of completion of the program revealed that the program was valuable to students and their motivation increased.

Automated brain tumor segmentation plays an important role in the diagnosis and prognosis of the patient. In addition, features from the tumorous brain help in predicting patients' overall survival. The main focus of this paper is to segment tumor from BRATS 2018 benchmark dataset and use age, shape and volumetric features to predict overall survival of patients. The random forest classifier achieves overall survival accuracy of 59% on the test dataset and 67% on the dataset with resection status as gross total resection. The proposed approach uses fewer features but achieves better accuracy than state-of-the-art methods.

A person is commonly described by attributes like height, build, cloth color, cloth type, and gender. Such attributes are known as soft biometrics. They bridge the semantic gap between human description and person retrieval in surveillance video. The paper proposes a deep learning-based linear filtering approach for person retrieval using height, cloth color, and gender. The proposed approach uses Mask R-CNN for pixel-wise person segmentation. It removes background clutter and provides precise boundary around the person. Color and gender models are fine-tuned using AlexNet and the algorithm is tested on SoftBioSearch dataset. It achieves good accuracy for person retrieval using the semantic query in challenging conditions.

The primary focus of this paper is to detect disease and estimate its stage for a cotton plant using images. Most disease symptoms are reflected on the cotton leaf. Unlike earlier approaches, the novelty of the proposal lies in processing images captured under uncontrolled conditions in the field using normal or a mobile phone camera by an untrained person. Such field images have a cluttered background making leaf segmentation very challenging. The proposed work use two cascaded classifiers. Using local statistical features, first classifier segments leaf from the background. Then using hue and luminance from HSV colour space another classifier is trained to detect disease and find its stage. The developed algorithm is a generalised as it can be applied for any disease. However as a showcase, we detect Grey Mildew, widely prevalent fungal disease in North Gujarat, India.

This paper addresses the issue of headlight intensity to alleviate glare and blinding during night for drivers. Many factors are considered when analyzing automobile transportation in order to increase safety. One of the most prominent factors for night-time travel is temporary blindness due to elevated headlight intensity. This is particularly prominent on single lane roads. While headlight intensity provides better visual acuity, it inversely affects oncoming traffic. This problem is compounded when both drivers are using a higher headlight intensity setting. Also, higher speed due to decreased traffic levels at night increases the severity of accidents. In order to eliminate accidents due to temporary driver blindness, a fuzzy controller is designed based on the data captured using a wireless sensor network (WSN). Low latency allows quicker headlight intensity adjustment to minimize temporary blindness. Multiple attributes are taken into consideration for controller design. The results show that controller output is nearly instantaneous and generates control signal continuously.

Superresolution is an algorithmic approach, for constructing high resolution de-noised image from its low resolution and noisier version. A new method to address the problem of copyright violation for super resolution is presented in this paper. The goal is to design an improved watermarking technique, while minimizing distortion in the super resolved image. The approach employs, fuzzy logic to build the perceptual mask, embeds watermark in the low frequency coefficients for robustness with edge preservation and use neural network at the receiver. Novelty lies in providing copyright protection jointly to the low resolution and the super resolved images. The distortion due to watermark insertion is compensated by: 1. use of fuzzy perceptual mask tuned to human visual system, 2. use of trained neural network estimator during watermark extraction, 3. utilize image degradation model during watermark extraction. Effectiveness of the proposed approach is shown by conducting the experiments on natural images and comparing it with the state of the art techniques.