search

ChaojianYu / Hierarchical-Bilinear-Pooling

Implementation for <Hierarchical Bilinear Pooling for Fine-Grained Visual Recognition> in ECCV'18.
Other
103 stars 23 forks source link

do you train on resent models? #2

Open zimenglan-sysu-512 opened 6 years ago

zimenglan-sysu-512 commented 6 years ago

Hi @ChaojianYu

Bilinear pooling works well on VGG16, but do you ever train on resnet model? like resnet50.

Thanks.

guyibang commented 5 years ago

i'd like to know too

Ylexx commented 5 years ago

@guyibang I did a simple implementation on resnet, if you are interested, you can refer to it. https://github.com/Ylexx/Hierarchical-Bilinear-Pooling_Resnet_pytorch

shiyuanyin commented 5 years ago

@ChaojianYu Hi ,thank you for your paper work ,is cool. but I have some question to ask for your help.

  1. I want to embed this structure in other model structures. If I change the model and dataset I also need to try it？？dimensions=512 1024 2048 4096 8192. Could you give me some advice on getting the best performance of the model?

Fig. 2.Classification accuracy on the CUB dataset. Comparison of general FactorizedBilinear Pooling (FBP), Cross-layer Bilinear Pooling (CBP) and Hierarchical BilinearPooling (HBP) with various projection dimensions

  1. I think your visilization is cool ,I also want to realise it.Could you provide me with relevant information and code?

4.4 Qualitative Visualization To better understand our model, we visualize the model response of differentlayers in our fine-tuned network on different datasets.

Fig. 3.Visualization of model response of different layers on the CUB, Cars and Air-craft datasets. It can be seen that our model tend to ignore features in the clutteredbackground and focus on the most discriminative parts of object