Open zzdang opened 6 years ago
Get the same question here.
Thank you for your attention! I am sorry that I was too busy last month and reply so late. The major difference is that I use gluon rather than caffe.
Some tricks to improve the performance are ignored. For example, multiply the regression term to 10 times in RCNN branch. It is just a project for fun, but not a serious re-implementation.
Hi ,your project is cool. I plan to write a faster rcnn project with gluon. I want to use your code for reference,could you list the differents between you codes and official faster rcnn?