MagicSource

https://www.mediafire.com/file/93l5l22mgj7dz59/voc_mb2_ssdlite_deploy.prototxt/file https://www.mediafire.com/file/16d5e532b4hs7xx/voc_mb2_ssdlite_deploy.caffemodel/file

Above prototxt and caffemodel train on VOC for 120000 iters, but can not got any results

no you can not do that

Well, if I am not using mege_bn, directly using iter 73000 caffemodel, the result seems correct:  Does the merge bn scripts wrong?

@ganler @luomai Yes, I mean training APi migration from 1.1x to tensorflow 2.0

May I ask why maskiou_head trends to get lower confidence?

@echarlaix thanks for your reply. What I am more interested is that if you have any experience on quantize a huge GPT2 model like 7.6GB in onnx size? I managed...

@alar0330 Will u opensource it? Looking forward for your speed and visualization results.

Hello, guys. I saw there so many languages, but just not Chinese, does there any plan to support Chinese?

Does it avaiable for now