search

zengyu714 / segmentation

Use several neural networks to segment IVDs.
15 stars 11 forks source link

Performance of refineNet #1

Open saeedizadi opened 7 years ago

saeedizadi commented 7 years ago

Hi there, Hope you are doing well,

I found that you've imlemented RefineNet in TF which sounds so interesting to me. Does your refineNet model works well? I mean, does it outperform other models in terms of accuracy?

Thanks Saeed

zengyu714 commented 7 years ago

Hey,

Refine block actually performs better when fusion 2 or more inputs than directly concat or add. But I didn't verify the results reported in its paper. For now, I just use the block rather than the whole framework to segment medical images.

Also, the TF implements are not very optimized and robust, so they are only suggestions : P

2017-07-26 6:49 GMT+08:00, Saeed Izadi notifications@github.com:

Hi there, Hope you are doing well,

I found that you've imlemented RefineNet in TF which sounds so interesting to me. Does your refineNet model works well? I mean, does it outperform other models in terms of accuracy?

Thanks Saeed

-- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/zengyu714/segmentation/issues/1