search

ecobost / cnn4brca

Using Convolutional Neural Networks (CNN) for Semantic Segmentation of Breast Cancer Lesions (BRCA). Master's thesis
75 stars 51 forks source link

how to run your code? #1

Open PapaMadeleine2022 opened 7 years ago

PapaMadeleine2022 commented 7 years ago

hello, I am interest at your code. But how to run your code? can you give some advice？ beside, can you provide some trainning data?

ecobost commented 7 years ago

I added some code on how to run in the README.md: https://github.com/ecobost/cnn4brca

PapaMadeleine2022 commented 7 years ago

@ecobost thank your response. can you provide a good pre-trained model? thank you very much!

Maya1994 commented 7 years ago 
@IvyGongoogle did u get the database ? i asked them but they didn't reply !
PapaMadeleine2022 commented 7 years ago

@Maya1994 sorry, I have not got it

Maya1994 commented 7 years ago

did u try the code with other dataBase ? can we try it with the MIAS data base ??

ecobost commented 7 years ago

@Maya1994 Hi, people at BCDR do take some time to answer, I have not tried to run this code in a different database but it should be possible although with some modifications

Maya1994 commented 7 years ago

@ecobost here is an example of mammograms that I have, as I told you i'm interested with microcalcification detection, image11 image11_label

You asked me before to generate masks first , but how can I do it ? as u said in ur previous comment : " what createMask.m does is take an outline (a series of x,y coordinates) and creates the contour of the lesion from it" but i don't have this x and y !!!

Many Thanks,

ecobost commented 7 years ago

Do you have any ground truth of where the lesions are? that has to be provided to the network as the label for training otherwise it will have no examples to learn from. If you do have some ground truth as masks, outlines or any other form, then you would have to change it to the form used by the code (with 0 for background, 127 for normal tissue and 255 for the lesion/positive class). This should probably come with the dataset (and are usually provided by trained radiologists), if you do want to label the images yourself you could try using labelme or sloth (I haven't used them but they seem relatively easy to use)

EngNala commented 7 years ago

mias data base provide x,y image-coordinates of centre of abnormality.

shubhamnatani commented 6 years ago

hey are you run the whole code.? the code gives some error when i run . please help if you know.?

shubhamnatani commented 6 years ago

@ecobost sir the code gives an error. can you help me out.?