search

alexgkendall / caffe-segnet

Implementation of SegNet: A Deep Convolutional Encoder-Decoder Architecture for Semantic Pixel-Wise Labelling
http://mi.eng.cam.ac.uk/projects/segnet/
Other
1.08k stars 452 forks source link

Running segnet on medical images #105

Open sajjo79 opened 7 years ago

sajjo79 commented 7 years ago

Hi, I am running segnet on medical images and in this context, i have few questions.

  1. I am giving grayscale images as input. The input layer is converting it into 3D images as it does for other RGB images. Is this input correct to the network.
  2. I have 5 classes in my images and results produced are 5 maps ( i have changed parameters accordingly). I have deleted 6 weights and kept 5 top weights as given in train.prototxt The feature maps produced are not too clear and all approximately seems to be similar. Do i need to update the weight parameters or other parameters.
  3. There is no code to quantify the results produced. Is there any.

Best.

ArunJ1 commented 5 years ago

Did you find any answer for your second question? I am also stuck whether i need to calculate my own weight or use existing?