Transportation is considered the fundamental pillarof economic growth in any society. Still, inherent issues likeaccidents, higher fuel consumption, and pollution have pavedthe way for the rise of the intelligent transportation system(ITS), which enables safety and improvement in the existingtransportation system. ITS helps the massive collection of datafrom multiple sources, and this big data needs immediate processing for ascertaining the events. However, prediction accuracycontinues to be low because of the trade-off of accuracy with theimmediate detection of an event, and this limits the performanceof ITS. This paper addresses this issue by proposing a novel IoT-enabled distributed context-aware Fog-cloud architecture, whichimproves the prediction accuracy by utilizing a hybrid CNN(Convolutional Neural Network) deep learning (DL) approach.Each vehicle in the system only has a track of the local knowledge.The nearby fog nodes are enabled to know the global eventsthrough incremental federated learning, which gets updatedcontinuously back and forth with fog and cloud. Experimentsdemonstrated in the paper clearly show that the modified versionof VGGNet for the CNN model outperforms RGB images,delivering an accuracy of more than 95%, which is 3% moreaccurate than the LeNet while using RGB images as input