I have trained the model for 40K iterations, the accuracy for the training process is about 90%, but when I test the model use the following command, the mAp is about 30%. I guess if I have the wrong testing process. could you help me?
"./tools/test_net.py --gpu 0 --def models/pascal_voc/ResNet-101/rfcn_end2end/test_agnostic.prototxt --net /home/who/rfcn/py-R-FCN/output/rfcn_end2end_ohem/voc_0712_trainval/resnet101_rfcn_ohem_iter_40000.caffemodel --imdb voc_0712_test --cfg experiments/cfgs/rfcn_end2end_ohem.yml 2>&1 | tee /home/who/rfcn/py-R-FCN/it40000-test.log"
I have trained the model for 40K iterations, the accuracy for the training process is about 90%, but when I test the model use the following command, the mAp is about 30%. I guess if I have the wrong testing process. could you help me? "./tools/test_net.py --gpu 0 --def models/pascal_voc/ResNet-101/rfcn_end2end/test_agnostic.prototxt --net /home/who/rfcn/py-R-FCN/output/rfcn_end2end_ohem/voc_0712_trainval/resnet101_rfcn_ohem_iter_40000.caffemodel --imdb voc_0712_test --cfg experiments/cfgs/rfcn_end2end_ohem.yml 2>&1 | tee /home/who/rfcn/py-R-FCN/it40000-test.log"