Browsing by Author "Chang, Ya-Ching"

Now showing 1 - 2 of 2

DEEP LEARNING-BASED HUMAN ACTIVITY ANALYSIS FOR AERIAL IMAGES
(2017-01-01)
Wang, Han-Yang
;
Chang, Ya-Ching
;
Hsieh, Yi-Yu
;
Chen, Hua-Tsung
;
Chuang, Jen-Hui
Due to the advantages of high mobility and the ability to fly in the sky, drone has inspired more and more applications in recent years. On the other hand, deep learning-based human activity analysis is an important research topic in security surveillance; however, there are few research works on such analysis with aerial images so far. Because of perspective projection, people in aerial images look tilted, which would degrade the performance of human activity analysis. In order to cope with the issue of perspective projection for aerial images, we modify the CNN architecture of a state-ofthe-art object detection method, YOLOv2 [12], and build an aerial image dataset with a drone for new model training. Finally, a post -processing method is proposed to classify the pose of a detected person as normal or abnormal, so that the task of human activity analysis with aerial images can be accomplished.
PEDESTRIAN DETECTION IN AERIAL IMAGES USING VANISHING POINT TRANSFORMATION AND DEEP LEARNING
(2018-01-01)
Chang, Ya-Ching
;
Chen, Hua-Tsung
;
Chuang, Jen-Hui
;
Liao, I-Chun
;
資訊工程學系
Drones are well-liked nowadays. However, deep learning models for object detection still cannot have high detection rates for pedestrians in aerial images even though they already show high precision on PASCAL VOC 2007. The main challenges of aerial image analysis include: (i) the size of an object in aerial images can be very small, and (ii) the objects in aerial images are tilted outward due to perspective projection deformation, which make the pedestrians hard to recognize in aerial images. In this paper, we utilize image partition and vanishing point transformation to overcome the above challenges. Experimental results demonstrate that such pre-processing methods can indeed increase the detection rates significantly for some deep learning models.