high fps but slow detection of traffic signs

Question

i used about 39k images from the german traffic signs recognition benchmark dataset to train my yolov5 model. i used the command

!python train.py --img 416 --data data.yaml --cfg yolov5s.yaml --batch-size 10 --epochs 100 --name Model

to train the model

am I missing something?

when im training with smaller dataset with about 1000k images it is much faster and can detect multiple traffic signs..

pls help