RAPiD: Rotation-Aware People Detection in Overhead Fisheye Images

@article{Duan2020RAPiDRP,
  title={RAPiD: Rotation-Aware People Detection in Overhead Fisheye Images},
  author={Zhihao Duan and M. Tezcan and Hayato Nakamura and Prakash Ishwar and Janusz Konrad},
  journal={2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)},
  year={2020},
  pages={2700-2709}
}
  • Zhihao Duan, M. Tezcan, +2 authors J. Konrad
  • Published 2020
  • Computer Science
  • 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
Recent methods for people detection in overhead, fisheye images either use radially-aligned bounding boxes to represent people, assuming people always appear along image radius or require significant pre-/post-processing which radically increases computational complexity. In this work, we develop an end-to-end rotation-aware people detection method, named RAPiD, that detects people using arbitrarily-oriented bounding boxes. Our fully-convolutional neural network directly regresses the angle of… Expand
Key strategies for synthetic data generation for training intelligent systems based on people detection from omnidirectional cameras
TLDR
This work generates different synthetic training datasets with particular features and uses them to train a DNN for people detection in large spaces using omnidirectional cameras, and shows that proper synthetic data minimize the domain gap. Expand
Deep authoring - an AI Tool set for creating immersive MultiMedia experiences
TLDR
A fully automated 360° video processing pipeline using a hierarchical combination of Artificial Intelligence (AI) modules to create immersive volumetric XR experiences and a detailed review of AI-based methods to address these problems is provided. Expand
Segmentation-Based Bounding Box Generation for Omnidirectional Pedestrian Detection
TLDR
A segmentation-based bounding box generation method for omnidirectional pedestrian detection, which enables detectors to tightly fit bounding boxes to pedestrians without omniddirectional images for training, and develops a pseudo-fisheye distortion augmentation method, which further enhances the performance. Expand
Unsupervised Domain Adaptation from Synthetic to Real Images for Anchorless Object Detection
TLDR
This paper implements unsupervised domain adaptation (UDA) methods on an anchorless object detector to adjust two UDA methods, viz., entropy minimization and maximum squares loss, originally developed for segmentation, to object detection. Expand

References

SHOWING 1-10 OF 38 REFERENCES
Supervised People Counting Using An Overhead Fisheye Camera
TLDR
Two supervised methods for people counting using an overhead fisheye camera are proposed that reduce the people counting MAE of two natural benchmarks by over 60%. Expand
People detection in top-view fisheye imaging
  • Oded Krams, N. Kiryati
  • Computer Science
  • 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS)
  • 2017
TLDR
A convenient purposive fisheye-camera calibration process is introduced, optimized for subsequent people detection performance, by applying a tunable f isheye camera model to project the fis heye image plane onto a unit half sphere, followed by the stereographic map projection. Expand
Improved Person Detection on Omnidirectional Images with Non-maxima Supression
TLDR
A person detector on omnidirectional images, an accurate method to generate minimal enclosing rectangles of persons using YOLOv2 to fish-eye images by fine-tuning overlapping bounding boxes by three approaches: non-maximum suppression, softNonmaximum suppression and soft non- maximum suppression with Gaussian smoothing. Expand
Template-based people detection using a single downward-viewing fisheye camera
TLDR
A new algorithm for detecting people using a single downward-viewing fisheye camera that is both practical to realistic environments and efficient enough for real-time applications is proposed. Expand
You Only Look Once: Unified, Real-Time Object Detection
TLDR
Compared to state-of-the-art detection systems, YOLO makes more localization errors but is less likely to predict false positives on background, and outperforms other detection methods, including DPM and R-CNN, when generalizing from natural images to other domains like artwork. Expand
Learning RoI Transformer for Oriented Object Detection in Aerial Images
TLDR
The core idea of RoI Transformer is to apply spatial transformations on RoIs and learn the transformation parameters under the supervision of oriented bounding box (OBB) annotations. Expand
SSD: Single Shot MultiBox Detector
TLDR
The approach, named SSD, discretizes the output space of bounding boxes into a set of default boxes over different aspect ratios and scales per feature map location, which makes SSD easy to train and straightforward to integrate into systems that require a detection component. Expand
FCOS: Fully Convolutional One-Stage Object Detection
TLDR
For the first time, a much simpler and flexible detection framework achieving improved detection accuracy is demonstrated, and it is hoped that the proposed FCOS framework can serve as a simple and strong alternative for many other instance-level tasks. Expand
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
TLDR
This work introduces a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals and further merge RPN and Fast R-CNN into a single network by sharing their convolutionAL features. Expand
Omnidirectional Pedestrian Detection by Rotation Invariant Training
TLDR
A rotation invariant training method, which only uses randomly rotated perspective images without any additional annotation for omnidirectional pedestrian detection, which achieved a state-of-the-art performance on four public benchmarks. Expand
...
1
2
3
4
...