-
-
Save ravnoor/a8d26c485cd39c1d9dd21af7c27ac232 to your computer and use it in GitHub Desktop.
V-Net in Keras and tensorflow
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| { | |
| "cells": [ | |
| { | |
| "cell_type": "code", | |
| "execution_count": 1, | |
| "metadata": { | |
| "collapsed": true | |
| }, | |
| "outputs": [], | |
| "source": [ | |
| "from os import listdir\n", | |
| "import SimpleITK\n", | |
| "import os.path\n", | |
| "import pickle\n", | |
| "import numpy" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 2, | |
| "metadata": { | |
| "collapsed": true | |
| }, | |
| "outputs": [], | |
| "source": [ | |
| "volSize = numpy.array((128,128,64), numpy.int32)\n", | |
| "dstRes = numpy.array((1,1,1.5))\n", | |
| "normDir = False\n", | |
| "method = SimpleITK.sitkLinear" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 3, | |
| "metadata": { | |
| "collapsed": true | |
| }, | |
| "outputs": [], | |
| "source": [ | |
| "def process_scan(scan):\n", | |
| " ret = numpy.zeros(volSize, dtype=numpy.float32)\n", | |
| " factor = numpy.asarray(scan.GetSpacing()) / dstRes\n", | |
| "\n", | |
| " factorSize = numpy.asarray(scan.GetSize() * factor, dtype=numpy.float)\n", | |
| "\n", | |
| " newSize = numpy.max([factorSize, volSize], axis=0)\n", | |
| "\n", | |
| " newSize = newSize.astype(dtype=numpy.int32)\n", | |
| "\n", | |
| " T=SimpleITK.AffineTransform(3)\n", | |
| " T.SetMatrix(scan.GetDirection())\n", | |
| "\n", | |
| " resampler = SimpleITK.ResampleImageFilter()\n", | |
| " resampler.SetReferenceImage(scan)\n", | |
| " resampler.SetOutputSpacing(dstRes)\n", | |
| " resampler.SetSize(newSize.tolist())\n", | |
| " resampler.SetInterpolator(method)\n", | |
| " if normDir:\n", | |
| " resampler.SetTransform(T.GetInverse())\n", | |
| "\n", | |
| " imgResampled = resampler.Execute(scan)\n", | |
| "\n", | |
| "\n", | |
| " imgCentroid = numpy.asarray(newSize, dtype=numpy.float) / 2.0\n", | |
| "\n", | |
| " imgStartPx = (imgCentroid - numpy.array(volSize) / 2.0).astype(dtype=int)\n", | |
| "\n", | |
| " regionExtractor = SimpleITK.RegionOfInterestImageFilter()\n", | |
| " regionExtractor.SetSize(volSize.astype(dtype=numpy.int32).tolist())\n", | |
| " regionExtractor.SetIndex(imgStartPx.tolist())\n", | |
| "\n", | |
| " imgResampledCropped = regionExtractor.Execute(imgResampled)\n", | |
| "\n", | |
| " return numpy.transpose(\n", | |
| " SimpleITK.GetArrayFromImage(imgResampledCropped).astype(dtype=numpy.float),\n", | |
| " [2, 1, 0]\n", | |
| " )" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 4, | |
| "metadata": { | |
| "collapsed": false | |
| }, | |
| "outputs": [], | |
| "source": [ | |
| "def iterate_folder(folder):\n", | |
| " for filename in sorted(listdir(folder)):\n", | |
| " absolute_filename = os.path.join(folder, filename)\n", | |
| " segmentation_absolute_filename = absolute_filename[:-4] + '_segmentation.mhd'\n", | |
| " if filename.endswith('.mhd') and os.path.exists(segmentation_absolute_filename):\n", | |
| " yield absolute_filename, segmentation_absolute_filename" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 5, | |
| "metadata": { | |
| "collapsed": true | |
| }, | |
| "outputs": [], | |
| "source": [ | |
| "def load_data(folder):\n", | |
| " input_filenames, label_filenames = zip(*list(iterate_folder(folder)))\n", | |
| " \n", | |
| " X = numpy.array([process_scan(SimpleITK.ReadImage(f)) for f in input_filenames])\n", | |
| " y = numpy.array([process_scan(SimpleITK.ReadImage(f)) for f in label_filenames])\n", | |
| " \n", | |
| " return X, y > 0.5" | |
| ] | |
| }, | |
| { | |
| "cell_type": "code", | |
| "execution_count": 6, | |
| "metadata": { | |
| "collapsed": false | |
| }, | |
| "outputs": [], | |
| "source": [ | |
| "X, y = load_data('../data/PROMISE2012/train/')\n", | |
| "X.shape, y.shape, y.mean() \n", | |
| "\n", | |
| "with open('../data/PROMISE2012/train_data.p3', 'wb') as f:\n", | |
| " pickle.dump([X, y], f)" | |
| ] | |
| } | |
| ], | |
| "metadata": { | |
| "kernelspec": { | |
| "display_name": "Python 3", | |
| "language": "python", | |
| "name": "python3" | |
| }, | |
| "language_info": { | |
| "codemirror_mode": { | |
| "name": "ipython", | |
| "version": 3 | |
| }, | |
| "file_extension": ".py", | |
| "mimetype": "text/x-python", | |
| "name": "python", | |
| "nbconvert_exporter": "python", | |
| "pygments_lexer": "ipython3", | |
| "version": "3.5.1" | |
| } | |
| }, | |
| "nbformat": 4, | |
| "nbformat_minor": 1 | |
| } |
It's the PROMISE12 dataset.
Hello,
Thanks for your implementation.
With;
model.fit(X, y, batch_size=4, epochs=5, verbose=1)
I have following exception;
InvalidArgumentError (see above for traceback): Incompatible shapes: [8388608] vs. [1048576]
- I am using same dataset with yours.
- Load data.ipynb works fine.
- I have tried with different batch_size, i.e. 4,5,10,50
- The model's summary is exactly same with yours.
I just wonder, do you have any idea, why I am getting this exception?
Thank you, regards.
- Hakan
hello,
I run U-Net using dice loss, but the predicted images are all white. Do you know what's wrong?
def dice_coef(y_true, y_pred):
smooth = 1
y_true_f = K.flatten(y_true)
y_pred_f = K.flatten(y_pred)
intersection = K.sum(y_true_f * y_pred_f)
return (2. * intersection +smooth) / (K.sum(y_true_f) + K.sum(y_pred_f) +smooth)
def dice_coef_loss(y_true, y_pred):
print("dice loss")
return 1-dice_coef(y_true, y_pred)
....
model.compile(optimizer = Adam(lr = 1e-5), loss = dice_coef_loss, metrics = ['accuracy'])
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
What dataset was your model used on?