search

foolwood / SiamMask

[CVPR2019] Fast Online Object Tracking and Segmentation: A Unifying Approach
http://www.robots.ox.ac.uk/~qwang/SiamMask
MIT License
3.47k stars 819 forks source link

The structure of SiamRPN++ you implement is different from that in Paper. #55

Closed TeslaHua closed 5 years ago

TeslaHua commented 5 years ago

Hi,Qiang Thks for your contribution!        I trained and tested the SiamRPN++ network you implemented and studied the code. Some questions now:

  1. The output model structure (only layer3 in the custom version of resnet50, why?) is different from the SiamRPN++ model in pysot.

  2. SiamRPN++ indicates in paper that the feature fusion of layer2_RPN, layer3_RPN and layer4_RPN in ResNet50 is not found in your code.

    look forward to your reply.

luoshanwei commented 5 years ago

I have the same questions

foolwood commented 5 years ago

because in my paper, I only compare with layer 3 version of SiamRPN. (SiamMask only use ResNet feature from layer 3).

If you want to reimplement the original version of SiamRPN++, please refer pysot.