Open hmtrii opened 3 years ago
Obtaining the grid-shaped feature can be equivalent to obtaining the new grid-shaped boxes from the original bounding boxes of left/right lung, you can refer to the code.txt
Hi @jeline0110 , could you public the training code for baseline experimental results in the paper. Thank you.
Hi @hmtrii @jeline0110 , sorry to disturb you. I have some confusion about the Anatomical Structure Relation Module in the SAR-net paper.
The output of pretrained PSPNet is a vector (shape <anatomical_part_nums, h, w>). Based this vector, I notice that author uses coordinate difference to quantify spatial relation with each anatomical part. I don't how to get the coordinate of each annatomical part. Is it to find the circumscribed rectangle for every organ on the binary mask image? Could you help me solve this confusion?
Thank you!
Hi @hmtrii @jeline0110 , sorry to disturb you. I have some confusion about the Anatomical Structure Relation Module in the SAR-net paper. The output of pretrained PSPNet is a vector (shape <anatomical_part_nums, h, w>). Based this vector, I notice that author uses coordinate difference to quantify spatial relation with each anatomical part. I don't how to get the coordinate of each annatomical part. Is it to find the circumscribed rectangle for every organ on the binary mask image? Could you help me solve this confusion?
Thank you!
Yes, we find the circumscribed rectangle for every organ on the binary mask image.
Hi @hmtrii @jeline0110 , sorry to disturb you. I have some confusion about the Anatomical Structure Relation Module in the SAR-net paper. The output of pretrained PSPNet is a vector (shape <anatomical_part_nums, h, w>). Based this vector, I notice that author uses coordinate difference to quantify spatial relation with each anatomical part. I don't how to get the coordinate of each annatomical part. Is it to find the circumscribed rectangle for every organ on the binary mask image? Could you help me solve this confusion?
Thank you!
Yes, we find the circumscribed rectangle for every organ on the binary mask image.
I got it! Thanks!
Hi, I'm having a problem when trying to implement your proposed network in A Structure-Aware Relation Network for Thoracic Diseases Detection and Segmentation paper. In session 3.2 Contextual Relation Module of the paper, I don't understand clearly how you create the grid-shaped feature. Could you explain to me the operation you did to obtain the grid-shaped feature?
Thank you a lot. ❤️