search

qjadud1994 / RetinaNet_tensorflow

For easier and more readable tensorflow codes
MIT License
13 stars 4 forks source link

first one? #1

Open 1453042287 opened 5 years ago

1453042287 commented 5 years ago

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

qjadud1994 commented 5 years ago

Since the backbone(Resnet) uses imagenet pretrained weight, learning proceeds steadily.

If you do not use pre-trained weights, your learning will not work.