vehicle_detection

The website is used to identify vehicles, assess and visualize traffic conditions in Ho Chi Minh City, Vietnam

Data

Data is get from real time camera data from Cong Thong Tin Giao Thong Thanh Pho Ho Chi Minh

Data Summary:

• Num of picture: 2586
• Num of object: 16753
• Num of class: 4 (motorcycle, car, bus, truck)
• Resolution of picture: 800x450

Class	motorcycle	car	bus	truck
Amount	12912	2085	411	1345

Data Augmentation

• Random Mirror
• Photometric Distortion
• Random Crop
• Random Expand
• Subtract Means

Model

Using SSD Architecture with VGG16 backbone. Pretrained model from qfgaohao - pytorch-ssd

mAP Accuracy

Application

Web development with Streamlit

Web link here

Feature:

• Detect Vehicles
  • From realtime data
  • From image
  • From video
• Visualize Realtime Information With Charts
• Notifications When Traffic Jams With Slack

How to use

• Realtime data: Choose Run on Live Camera, choose camera name and set threshold.

  Note: The data is updated each 15 seconds. It's because of camera from Cong Thong Tin Giao Thong Thanh Pho Ho Chi Minh

• Image data: Choose Run on Image and upload your image

• Video data: Choose Run on Video and upload your video. You can check Record Video to save output video.