Vehicle Classification in Inclement Weather Conditions Using Deep Learning


  • Abdul Jabbar M
  • Pranesh K
  • Gowri J


RetinaNet, Double subnet network, Convolutional brain, Public regular hazy dataset.


In the past portion of 10 years, object discovery approaches given convolutional brain networks have been generally examined and effectively applied in numerous PC vision applications. In any case, identifying objects in harsh weather patterns stays a significant test in light of unfortunate permeability. In this paper, we address the item discovery issue within the sight of mist by presenting a novel double subnet network (DSNet) that can be prepared from start to finish and mutually learn three errands: permeability improvement, object grouping, and item restriction. DSNet achieves total execution improvement by including two subnetworks: the location subnet and the rebuilding subnet. We utilize RetinaNet as a spine organization (likewise called discovery subnet), which is liable for figuring out how to order and find objects. The reclamation subnet is planned by offering highlight extraction layers to the discovery subnet and embracing a component recuperation (FR) module for permeability upgrade. Exploratory outcomes show that our DSNet accomplished 50.84% mean normal accuracy (Guide) on an engineered hazy dataset that we created and 41.91% Guide on a public regular hazy dataset (Hazy Driving dataset), outflanking many cutting edge object locators and blend models among tease and discovery strategies while keeping a fast.


Download data is not yet available.




How to Cite

Abdul Jabbar M, Pranesh K, & Gowri J. (2023). Vehicle Classification in Inclement Weather Conditions Using Deep Learning. International Journal of Progressive Research in Science and Engineering, 4(5), 66–69. Retrieved from