康行天下
@GuoxingYan I didn't met. the integer seems to be abnormal big.
@GuoxingYan I think it needs to be.
@licy5152 I wrote a python script [CascadeRCNN-demo.py](https://gist.github.com/makefile/6731ca0e311b6401681c15635bb97330) imitate the matlab code, you can modify it to use.
@GuoxingYan 你的网络问题吧
@PacteraKun The situation you encountered is unusual, check carefully.
@liulei01 My training loss of ship task doesn't decrease and keeps 8.0 ~ 9.0 between iterations of 5000 ~ 20000. I do as the README instruction and used the rbox_util.cpp.ship...
The app.py is an example script for detection by using Caffe or Darknet YOLO models. You need modify and write your own code for underlying detection functions.
I met the same issue with `-dr` flag. There may be a bug in the training of word2vec.
maybe learning rate is too large