Vision Based Pedestrian Traffic Counting Using Deep Learning Method
Yolov4 is a one stage detector to the vision-based object detection system. It is a predictive technique that provides faster and accurate results with minimal background errors. Object detection is a computer vision technique that performs to identify and locate objects within an image or video. In...
Saved in:
| Main Author: | |
|---|---|
| Format: | Undergraduates Project Papers |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/39914/1/EA18177_Ho_Thesis%20-%20Ho%20Chia%20Hui.pdf http://umpir.ump.edu.my/id/eprint/39914/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Yolov4 is a one stage detector to the vision-based object detection system. It is a predictive technique that provides faster and accurate results with minimal background errors. Object detection is a computer vision technique that performs to identify and locate objects within an image or video. In other word, object detection draws bounding boxes around these detected objects, which allow us to know where objects are in. One of the challenges of object detection is occlusion reduce the detection accuracy. The aim of this project is to detect and track the pedestrian even they are walk in group. The output of the bounding box is obtained after the input image passed through the Yolov4 network architecture. After that the threshold and non-maximum suppression (NMS) are applied to get the best bounding box. The counting function is done when after NMS. Score threshold is adjustable to observe which thresholds can get a better accuracy result in an image or video. The accuracy is obtained by applying the formula of TP, TN, FP and FN. The result shown that using score threshold of 0.35 can get higher accuracy which is 84.62%~100% after simulate. |
|---|