Developing Traffic Congestion Detection Model Using Deep Learning Approach: Case Study of Addis Ababa city Road

Traffic congestion is one of the most irritating problem that society face in their everyday
activity. In Ethiopia, the number of vehicles and pedestrians is increasing at a high rate
from time to time. Excessive numbers of traffic on roads and improper control of traffic
create traffic congestion. Uncontrolled traffic congestion hinders the transportation of
goods and commuters from place to place and increases the volume of carbon emitted into
the air. It can also either hampers or stagnates schedule, business, and commerce. Many
images and video processing approaches have been researched across the world to detect
traffic congestion. One such approach is that of using background and foreground
subtraction, average frame differencing, and convolutional neural network to detect traffic
congestion from video. From different literature review one-stage object detectors
identified as the best methods to detect traffic congestion with acceptable accuracy and
speed. In this study one-stage object detectors are used to detect traffic congestion from
video to improve speed problems that others researchers left as research gaps while using
two-stage object detectors. Dataset is prepared by extracting frames from different video
footage. The extracted frames were labeled manually as congested and uncongested. To
train, the models pre-trained Darknet weights were used. YOLOv3 and YOLOv5 model
used for experimentation. Accuracy and speed metrics used to evaluate the performance of
the models. A YOLOv3 and YOLOv5 model achieved 68.6%, and 62.3 % mean average
precision on a testing dataset respectively.