search

JinyuanLiu-CV / SegMiF

ICCV2023 | Multi-interactive Feature Learning and a Full-time Multi-modality Benchmark for Image Fusion and Segmentation
109 stars 11 forks source link

A beginner's sincere request #8

Open Liam-Merouin opened 1 year ago

Liam-Merouin commented 1 year ago

I'm sorry to bother you again. I have a few questions:

  1. How are image fusion networks and segmentation networks connected? Did he directly feed the generated fusion network feature map into the segmentation network?
  2. How is the segmentation feature map of the segmentation network fed back to HIA? I see that several modules in the code directly use segfeature parameters, that the first time when there is no segfeature he is how to run ah?

A beginner's sincere request

LiuZhu-CV commented 1 year ago

(1) The fusion and segmentation network are cascaded, where the ouput of fusion network is the input of segmentation network. (2) We utilize the pretrained backbone (VIT-b3 on Imagenet) to generate segfeatures in the first time.

Liam-Merouin commented 1 year ago

Hello, the data folder is empty except for the Mask folder It doesn't work when running train.py and test_fusion.py. Is the MFNET dataset used in your master code? Or the FMB you created yourself

Liam-Merouin commented 1 year ago

@LiuZhu-CV （＾ｖ＾）

LiuzhuForFun commented 1 year ago

I will upload these datasets tomorrow.

Liam-Merouin commented 1 year ago

Thank you, I hope you can share the link with me after uploading, because I did not see the new uploaded content in that Google cloud disk

LiuZhu-CV commented 1 year ago

Sorry I have uploaded the MFNet dataset at “https://drive.google.com/drive/folders/1MFTVd32-VNcpiFfNsu9Rw73YZJATPHA6”