Nayak, Satyajit (Valeo India Pvt. Ltd.), Patitapaban, Palo (Valeo India Private Limited), Gupta, Kwanit (Valeo India Pvt. Ltd.), Uttarkabat, Satarupa (Valeo India Pvt. Ltd.)

Graph-Based Two-Three Wheeler Classification in Unconstrained Indian Roads

Scheduled for presentation during the Poster Session "Transportation Data Analysis and Calibration" (FrAT17), Friday, September 27, 2024, 10:30−12:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on April 29, 2025

Abstract

Driven by the continuous advancements in the field of Advanced Driver Assistance Systems (ADAS), object detection and classification are crucial functionalities receiving significant attention in recent research. Mainly, Indian scenarios are highly dynamic due to the diverse and challenging driving conditions, including congested urban traffic, unpredictable road infrastructure, and varying weather patterns. This paper presents a unique method for classifying two and three-wheelers on the challenging Fine-Grained Vehicle Detection (FGVD) dataset from India. In the first stage, we employ the yolov7-tiny model to detect the objects in the image. In the second stage, we represent the detected object images in the graph domain and utilize a GNN architecture to classify objects within the image. A Spatial Graph Convolutional Network (SGCN) is employed in the second stage to classify the detected objects based on their spatial context. This approach leverages the strengths of SGCNs for classification, potentially leading to improved performance on the two, and three-wheelers of the FGVD dataset. We evaluate our method on the FGVD dataset and demonstrate its effectiveness in achieving object detection and classification.