训练之后，进行测试。map为什么只有0.0几呢，basenet为None是意味着不需要下载预训练权重吗

[fine2copyV]

When I trained with my own dataset, its loss converged, dropping from 15 to 0.5; then when I ran the test, the mAP was all zero