Academia.eduAcademia.edu

Figure 15 – uploaded by Swadesh Maurya

See full PDFdownloadDownload figure
Fig. 13: Another frame from online analysis of video [13]. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.

Figure 13 Another frame from online analysis of video [13]. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.

Related Figures (15)

Fig. 1: An example of a detected pedestrian. Images are best viewed in color.
Fig. 1: An example of a detected pedestrian. Images are best viewed in color.
Fig. 2: An example of a detected pedestrians when there ar many people together. Images are best viewed in color.
Fig. 2: An example of a detected pedestrians when there ar many people together. Images are best viewed in color.
Fig. 4: Division of the image in the different regions. The topmost R, is farthest from the vehicle, while the middle of Ry and R3z are the vulnerable regions. Dividing the images into these regions, is necessary since we are working with a single camera and there is no depth information. Size of the vulnerable road user is dependent on the distance from the vehicle. These regions allow us to find the degree of vulnerability accurately without information of distance from the vehicle. Fig. 3: Labels of the different regions in which we divide the image.
Fig. 4: Division of the image in the different regions. The topmost R, is farthest from the vehicle, while the middle of Ry and R3z are the vulnerable regions. Dividing the images into these regions, is necessary since we are working with a single camera and there is no depth information. Size of the vulnerable road user is dependent on the distance from the vehicle. These regions allow us to find the degree of vulnerability accurately without information of distance from the vehicle. Fig. 3: Labels of the different regions in which we divide the image.
Fig. 5: Distance vs Degree of Vulnerability
Fig. 5: Distance vs Degree of Vulnerability
Fig. 7: The same scene in Figure 6, after filtering. Images are best viewed in color. Fig. 6: Instance of an invalid detection. Images are best viewed in color.
Fig. 7: The same scene in Figure 6, after filtering. Images are best viewed in color. Fig. 6: Instance of an invalid detection. Images are best viewed in color.
Fig. 8: It can be seen that even though there are a large number of people in the scene, degree of vulnerability of the pedestrians has been accurately assigned based on the region and direction of movement. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable. It can be seen from the result images that our framework The Figures 8, 9, 10 and 11 are instances of the same video [12] across time. It can be seen that despite illumination variation and various dynamic occlusions in the scene, the pedestrians and cyclists are correctly detected and tracked and their vulnerability score is correctly computed by our proposed framework. Similarly, Figures 12, 13, 14, 15 and 16 show results of our proposed framework on voideo [13]
Fig. 8: It can be seen that even though there are a large number of people in the scene, degree of vulnerability of the pedestrians has been accurately assigned based on the region and direction of movement. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable. It can be seen from the result images that our framework The Figures 8, 9, 10 and 11 are instances of the same video [12] across time. It can be seen that despite illumination variation and various dynamic occlusions in the scene, the pedestrians and cyclists are correctly detected and tracked and their vulnerability score is correctly computed by our proposed framework. Similarly, Figures 12, 13, 14, 15 and 16 show results of our proposed framework on voideo [13]
Fig. 10: Inspite of illumination variations as the vehicle moves in the scene, accurate detection, tracking and coputation of degree of vulnerability of the pedestrians and cyclists is carried out through our proposed framework. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable. Fig. 9: Detection, tracking and degree of vulnerability of the pedestrians and cyclists. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.
Fig. 10: Inspite of illumination variations as the vehicle moves in the scene, accurate detection, tracking and coputation of degree of vulnerability of the pedestrians and cyclists is carried out through our proposed framework. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable. Fig. 9: Detection, tracking and degree of vulnerability of the pedestrians and cyclists. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.
Fig. 12: A frame from video [13], showing correct detection of VRUs and their vulnerability score by color of the bounding box. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable. Fig. 11: Accurate detection, tracking and degree of vulnera- bility of the VRUs from online analysis of video [12]. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.
Fig. 12: A frame from video [13], showing correct detection of VRUs and their vulnerability score by color of the bounding box. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable. Fig. 11: Accurate detection, tracking and degree of vulnera- bility of the VRUs from online analysis of video [12]. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.
Fig. 15: Pedestrians in region R, do not get detected, however, they are very far from the vehicle and therefore, their degree of vulnerability is 0. The objects of interest that are closer to the vehicle are correctly detected and their vulnerability is correctly computed. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable. Fig. 14: In spite of occlusion and a large number of vehicles in the scene, the VRUs are correctly detected, tracked and their degree of vulnerability is correctly shown by the color of the bounding box. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.
Fig. 15: Pedestrians in region R, do not get detected, however, they are very far from the vehicle and therefore, their degree of vulnerability is 0. The objects of interest that are closer to the vehicle are correctly detected and their vulnerability is correctly computed. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable. Fig. 14: In spite of occlusion and a large number of vehicles in the scene, the VRUs are correctly detected, tracked and their degree of vulnerability is correctly shown by the color of the bounding box. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.
Fig. 16: An instance showing accurate detection, tracking and degree of vulnerability of the pedestrians and cyclists. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.
Fig. 16: An instance showing accurate detection, tracking and degree of vulnerability of the pedestrians and cyclists. Red box shows highly vulnerable, yellow box shows predictability of vulnerability in the future and green box shows that the person is not vulnerable.
Related topics:
Computer VisionImage ProcessingMachine LearningIntelligent Transportation Systems
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved