• Gottemukula Hadassah Mercy
• Maggidi Prathyusha
• Pendoti Kavya
• Mr. Monis Tariq

Addressing real-world challenges in various smart city applications, such as parking occupancy detection, necessitates fine-tuning deep Neural Networks. For expansive parking areas, utilizing a high-placed cenital plane camera facilitates comprehensive monitoring with a single device. Leading object detection models like YOLO offer commendable precision and real-time speed, yet leveraging proprietary data provides significant room for customization beyond general-purpose datasets like COCO and ImageNet. This study proposes a modified, lightweight deep object detection model based on the YOLO-v5 architecture, capable of detecting objects of all sizes. Specifically, a multi-scale mechanism is introduced to learn discriminative features across different scales and automatically select the optimal scales for object detection, particularly vehicles. This multi-scale module reduces the number of trainable parameters compared to the original YOLO-v5 architecture. Experimental results demonstrate a substantial improvement in precision, with only a minor reduction in parameters from the YOLO-v5-S profile to our model. Moreover, the detection speed is reduced by inferring 30 fps compared to YOLO-v5-L/X profiles, while the performance in detecting tiny vehicles sees a significant 33% enhancement compared to the YOLO-v5-X profile.

Real-world challenges, Parking occupancy detection, Fine-tuning, Deep Neural Networks, High-placed cenital plane camera , Comprehensive monitoring, Object detection models, YOLO, Proprietary data, General-purpose datasets, COCO, ImageNet Modified model, Lightweight, YOLO-v5 architecture, Multi-scale mechanism, Optimal scales, Vehicle detection, Improvement Detection speed, Tiny vehicles

"T-YOLO: Tiny Vehicle Detection Based on YOLO and Multi-Scale Convolutional Neural Networks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.11, Issue 4, page no.p160-p163, April-2024, Available :http://www.jetir.org/papers/JETIR2404G21.pdf

"T-YOLO: Tiny Vehicle Detection Based on YOLO and Multi-Scale Convolutional Neural Networks", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.11, Issue 4, page no. ppp160-p163, April-2024, Available at : http://www.jetir.org/papers/JETIR2404G21.pdf