A5087267162- Advanced Neural Network Applications
- Video Surveillance and Tracking Methods
- Vehicle License Plate Recognition
- Autonomous Vehicle Technology and Safety
- Handwritten Text Recognition Techniques
- Robotics and Sensor-Based Localization
- Advanced Vision and Imaging
- Infrared Target Detection Methodologies
- Digital Transformation in Industry
- Maritime Navigation and Safety
- IoT and Edge/Fog Computing
- Traffic and Road Safety
- Software-Defined Networks and 5G
- Ergonomics and Musculoskeletal Disorders
- Infrastructure Maintenance and Monitoring
- Traffic Prediction and Management Techniques
- Remote Sensing and LiDAR Applications
- Optical measurement and interference techniques
- Visual Attention and Saliency Detection
- Generative Adversarial Networks and Image Synthesis
- Safety Warnings and Signage
- Advanced Data Processing Techniques
- Aerospace Engineering and Applications
- Human Pose and Action Recognition
- Noise Effects and Management
Brno University of Technology
2017-2023
In this paper, we focus on fine-grained recognition of vehicles mainly in traffic surveillance applications. We propose an approach that is orthogonal to recent advancements (automatic part discovery and bilinear pooling). addition, contrast other methods focused vehicles, do not limit ourselves a frontal/rear viewpoint, but allow the be seen from any viewpoint. Our based 3-D bounding boxes built around vehicles. The box can automatically constructed data. For scenarios where it possible use...
This work is focused on recognition of license plates in low resolution and quality images. We present a methodology for collection real world (non-synthetic) dataset plate images with ground truth transcriptions. Our approach to the based Convolutional Neural Network which holistically processes whole image, avoiding segmentation characters. Evaluation results multiple datasets show that our method significantly outperforms other free commercial solutions data. To enable further research...
In this paper, we focus on traffic camera calibration and a visual speed measurement from single monocular camera, which is an important task of surveillance. Existing methods addressing problem are difficult to compare due lack common data set with reliable ground truth. Therefore, it not clear how the in various aspects what factors affecting their performance. We captured new 18 full-HD videos, each around 1 hr long, at six different locations. Vehicles videos (20865 instances total)...
The 1 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">st</sup> Workshop on Maritime Computer Vision (MaCVi) 2023 focused maritime computer vision for Unmanned Aerial Vehicles (UAV) and Surface Vehicle (USV), organized several subchallenges in this domain: (i) UAV-based Object Detection, (ii) Mar-itime Tracking, (iii) USV-based Obstacle Segmentation (iv) Detection. were based the SeaDronesSee MODS benchmarks. This report summarizes main findings...
In this paper, we present a solution for automatic checkout in retail store as part of AI City Challenge 2022. We propose novel approach that uses the "removal" unwanted objects — case, body parts operating staff, which are localized and further removed from video by an image inpainting method. Afterwards, neural network detector can detect products with decreased detection false positive rate. A our is also ROI (the place where shown to system). reached 0.4167 F1-Score 0.3704 precision...
In this work, we propose a novel method for automatic camera calibration, mainly surveillance cameras. The calibration consists in observing objects on the ground plane of scene; our experiments, vehicles were used. However, any arbitrary rigid can be used instead, as verified by experiments with synthetic data. process uses convolutional neural network localisation landmarks observed scene and corresponding 3D positions localised - thus fine-grained classification detected image is done....
In our submission to the NVIDIA AI City Challenge 2020, we address problem of counting vehicles by their class at multiple intersections. Our solution is based on tracking principle using convolutional neural networks in detection and steps proposed method. We have achieved 6th place dataset part A Track 1 with score S1 Total = 0.8829, (mwRMSE 4.3616, Effectiveness 0.9094, Efficiency 0.8212). The was placed sixth overall ranking A.
In this paper, we explore the implementation of vehicle and pedestrian detection based on neural networks in a real-world application. We suggest changes to previously published method with respect capabilities low-powered devices, such as Nvidia Jetson platform. Our experimental evaluation shows that detectors are capable running 10.7 FPS TX2 can be used applications.
We explore the implementation of vehicle fine-grained type and color recognition based on neural networks in a real-world application. suggest changes to previously published method with respect capabilities low-powered devices, such as Nvidia Jetson. Experimental evaluation shows that accuracy MobileNet net slightly decreases compared ResNet-50 from 89.55% 86.13% while inference is 2.4× faster
The 1$^{\text{st}}$ Workshop on Maritime Computer Vision (MaCVi) 2023 focused maritime computer vision for Unmanned Aerial Vehicles (UAV) and Surface Vehicle (USV), organized several subchallenges in this domain: (i) UAV-based Object Detection, (ii) Tracking, (iii) USV-based Obstacle Segmentation (iv) Detection. were based the SeaDronesSee MODS benchmarks. This report summarizes main findings of individual introduces a new benchmark, called Detection v2, which extends previous benchmark by...
In our submission to the NVIDIA AI City Challenge, we address speed measurement of vehicles and vehicle re-identification. For both these tasks, use a calibration method based on extracted vanishing points. We detect track by CNN-based detector construct 3D bounding boxes for all vehicles. task, estimate from movement box in space using calibration. Our approach re-identification is extraction visual features "unpacked" images The are aggregated temporal domain obtain single feature...
In this paper, we explore the problem of license plate recognition in-the-wild (in meaning capturing data in unconstrained conditions, taken from arbitrary viewpoints and distances). We propose a method for automatic based on geometric alignment plates as preceding step holistic recognition. The is done by Convolutional Neural Network that estimates control points rectifying image following rectification formulated so whole process can be assembled into one computational graph contemporary...
Abstract Purpose Given the inconsistent application of various road markings on Czech rural roads, there is a question “How does marking in horizontal curves influence driving behaviour?” The study objective was to assess how centreline and edgelines behaviour. Methods To focus critical conditions, six secondary with radii below 200 m, were selected monitored before after marking. studied indicators average speed lateral position, which collected using trajectories detected calibrated video...
The goal of this work was to analyze the behavior vehicles on third-grade roads with and without horizontal lane markings small curvature (R ≤ 200m). are not frequented by many vehicles, therefore, a general short-term study would be able provide enough data. We used recording devices for long-term (weeks) traffic designed system analyzing trajectories employing computer vision. collected dataset at 6 distinct locations, containing 1 010 hours day-time video. In dataset, we tracked over 12...
A significant hurdle within any counting task is the variance in a scale of objects to be counted. While size changes some extent can induced by perspective distortion, more severe differences easily occur, e.g. case images taken drone from different elevations above ground. The aim our work overcome this issue leveraging only lightweight dot annotations and minimum level training supervision. We propose modification Stacked Hourglass network which enables model process multiple input scales...