search

ucbdrive / few-shot-object-detection

Implementations of few-shot object detection benchmarks
Apache License 2.0
1.1k stars 223 forks source link

mAP #174

Open salehnia opened 2 years ago

salehnia commented 2 years ago

Hi, Thanks for sharing your Code. The evaluation metric in your method is mAP. We know that the AP must be between 0 and 1. Considering that the result of executing your code on COCO in 10 shots is equal to 9.1. Does this mean that your model has a detection mAP about 0.09 in real?

Thank you

frankvp11 commented 1 year ago

I was just wondering about the evaluation metrics in general. It says bAP, but I'm only familiar with mAP. I'm also curious about the 9.1

muratbayrktr commented 1 year ago

Please refer to the paper https://arxiv.org/pdf/2003.06957.pdf. bAP stands for the baseAP meaning it measures the two-stage fine-tuning approach (TFA)'s performance over the base class after the fine tuning stages. The authors wanted to make sure the performance of the base class didn't vanish after the fine-tuning stage and that is why they included base classes' AP.