qjadud1994 / retinanet_tensorflow Goto Github PK

why the loss is so small? like this:
---------- open MonitoredTrainingSession
---------- run pretrain op
---------- run sync op
STEP : 0 TRAIN_TOTAL_LOSS : 1.21953082 TRAIN_LOC_LOSS : 0.08750351 TRAIN_CLS_LOSS : 1.13203
STEP : 50 TRAIN_TOTAL_LOSS : 1.22364414 TRAIN_LOC_LOSS : 0.09114464 TRAIN_CLS_LOSS : 1.13250
STEP : 100 TRAIN_TOTAL_LOSS : 1.21648896 TRAIN_LOC_LOSS : 0.08427663 TRAIN_CLS_LOSS : 1.13221

I only download the pretrained model file and rename it.
However, when I run the session, I get following error:
DataLossError (see above for traceback): Unable to open table file pretrained/se-resnet101-bn/se-resnet101-bn.pth: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?
What`s the problem?

论文中提到，在进行inference的时候，对FPN上的box进行了过滤@qjadud1994
inference involves simply forwarding an image through the network. To improve speed, we only decode box predictions from at most 1k top-scoring predictions per FPN level, after thresholding detector confidence at 0.05. The top predictions from all levels are merged and non-maximum suppression with a
threshold of 0.5 is applied to yield the final detections.