Undegraduate Minor Project on Vehicle Detection and Congestion Mapping
A Computer Vision based software to classify and count the total number of vehicles in the road and estimate the road occupancy with client-server based monitoring service. It employs YOLO CNN trained through transfer learning on self-constructed dataset of Nepali vehicles.
The program analyses a video surveillance footage, classifies the vehicles frame by frame and communicates the information to the client side through sockets using Django channels. This information is then used to visualize the current vehicles count and congestion in graphical format.
The data of the entire video is also accumulated to provide analytics like:
-How the road occupancy scenario changes with time
-How the vehicle count changes over time for each class of vehicle
-The contribution of each vehicle class to the overall congestion
8 classes of Nepali Vehicles are detected: Bike, Car, Taxi, Micro, Tempo, Bus, Pickup, Truck
---- Note, for the web sockets to function,additional dependencies are required ----
- Django Channels v1.0.2
- asgi_redis v1.0.0
- daphne v1.0.0
- asgiref v0.13
- redis v2.1
- autobahn v0.18
- twisted v16.2
---- Please note the latest version of Django Channels v2.x has significant code and architecture update and hence the code is incompatible with v2.x. Similarly, the other libraries are also compatible only with channels v1.x.
---- Install in the following order (because automatic dependency fetch from package installers like pip will download the latest but incompatible versions):
pip install twisted==16.2
pip install autobahn==0.18
pip install redis==2.1
pip install asgiref==0.13
pip install daphne==1.0.0
pip asgi_redis==1.0.0
pip install channels==1.0.2