Jerry Zhang
Jerry Zhang
@enry12 with mxnet 1.5 you can try amp. it automatically uses fp16, although slightly slower than manual fp16. You can check RCNN's script for amp training. For RCNN, we are...
https://github.com/dmlc/gluon-cv/blob/master/scripts/detection/faster_rcnn/train_faster_rcnn.py#L463
@JWarlock what is the batch size you are using? if batch size per gpu is less than 16 it might cause problems due to batch norm stats being not accurate....
In the latest SSD, the mAP is 77.2%, which is significantly better. There is also an tensorflow version here: https://github.com/balancap/SSD-Tensorflow
@catsdogone I tried the same thing and my activation is off, and cannot get the same accuracy. I also specified the scale parameter in scale layer to 1, and set...
Okay, I was able to achieve similar performance after changing the padding for the 1x7 and 7x1 filter to (0,3) and (3,0) instead of (1,2) and (2,1)
@nyyznyyz1991 I use pycaffe to generate the prototxt based on tensorflow. I cannot share it though.
@neobarney took me about 1 week.
@neobarney it should only take u 2-3 days, I spend half a week on figuring out why my activation does not match the original network.
@zmlmanly set scale to 1 on caffe should work