Different results compared with readme

Hi, I run ""CUDA_VISIBLE_DEVICES=0 python demo/inference_on_a_image.py -c groundingdino/config/GroundingDINO_SwinT_OGC.py -p weights/groundingdino_swint_ogc.pth -i /share/project/yanzhi/GroundingDINO/images/image.png -o ./ -t "The left lion", and the result detection is:

In your readme, only the left lion is grounded, and for me, all lions are grounded. Do you know how to ground the left lion only? Is there any settings I ignored or is there a specific input prompt? Thank you so much !