Pwkcoder

Training with my own dataset can achieve 0.95 mAP on FPN. Why on nas-fpn only have 0.3

Model usage

2

Since there is no masks information in my dataset, how can I use rfp for target frame detection instead of segmentation