BACKGROUND: With the growing number of people having vehicles in today's world, lots of problems occur with regards to a traffic violation, parking, etc. which arises the need for vehicle identification.
PROBLEM STATEMENT: Presently people are being used for handling the traffic and parking spaces. All these places have CCTV cameras but they are not proactive.
DATA: The dataset consists of total 786,702 images with 648,959 in the training dataset and 137,743 in the testing dataset which has been divided into six categories.
APPROACH: The data were first preprocessed where all the unnecessary images were removed and then trained with a various number of layers and parameter tuning like image size and a different number of layers.
SOLUTION: In this project, I have used six different categories of data for training namely two-wheeler, car, bus, truck, nonmotorized vehicle, and background.There is a total of three convolutional layers used to train the data and this resulted in an accuracy of 84 % on the test set. The pixels of the image have been reduced to 128 for the best result.
RESULTS: After the CNN was trained it was able to predict with an accuracy of 84% on the test dataset. This will be of great help once this model is used for handling of traffic or parking.