Sam20202/

UVH-26

  ArXiv   |  
 Models 






Models trained on UVH-26 deliver up to 31.5% higher mAP than COCO-pretrained baselines,
demonstrating significant gains in real-world performance for Indian traffic scenarios.







	
	Dataset Card for UVH-26 (Urban Vision Hackathon Dataset)




	
	Dataset Summary