Datasets
Models
Downloads: 1
[1fc123]:
/
Segthor_up.ipynb
1273 lines (1272 with data), 29.8 kB
{
"cells": [
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "FRm2ZqbPpr60"
},
"source": [
"# Image segmentation"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab": {},
"colab_type": "code",
"id": "KQbc5Xun-Vlf"
},
"source": [
"## Data preprocessing"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import numpy as np\n",
"import os\n",
"import time\n",
"import cv2\n",
"#import nibabel as nib\n",
"import pdb\n",
"from matplotlib import pyplot as plt\n",
"import nibabel as nib\n",
"from nibabel.testing import data_path"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"def truncated_range(img):\n",
" max_hu = 384\n",
" min_hu = -384\n",
" img[np.where(img > max_hu)] = max_hu\n",
" img[np.where(img < min_hu)] = min_hu\n",
" return (img - min_hu) / (max_hu - min_hu) * 255.\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"path = './train/'\n",
"save_path = './data/data_npy/'\n",
"\n",
"if not os.path.exists(save_path):\n",
" os.makedirs(save_path)\n",
"\n",
"files = os.listdir(path)\n",
"count = 0\n",
"print('begin processing data')\n",
"\n",
"means = []\n",
"stds = []\n",
"\n",
"\n",
"for i, volume in enumerate(files):\n",
" total_imgs = []\n",
" cur_file = os.path.join(path, volume)\n",
" print(i, cur_file)\n",
" cur_save_path = os.path.join(save_path, volume)\n",
" if not os.path.exists(cur_save_path):\n",
" os.makedirs(cur_save_path)\n",
" img = nib.load(os.path.join(cur_file, volume + '.nii'))\n",
" img = np.array(img.get_data())\n",
" label = nib.load(os.path.join(cur_file, 'GT.nii'))\n",
" label = np.array(label.get_data())\n",
" img = truncated_range(img)\n",
" for idx in range(img.shape[2]):\n",
" if idx == 0 or idx == img.shape[2] - 1:\n",
" continue\n",
" # 2.5D data, using adjacent 3 images\n",
" cur_img = img[:, :, idx - 1:idx + 2].astype('uint8')\n",
" total_imgs.append(cur_img)\n",
" cur_label = label[:, :, idx].astype('uint8')\n",
" count += 1\n",
" np.save(os.path.join(cur_save_path,volume + '_' + str(idx) + '_image.npy'), cur_img)\n",
" np.save(os.path.join(cur_save_path,volume + '_' + str(idx) + '_label.npy'), cur_label)\n",
" \n",
" total_imgs = np.stack(total_imgs, 3) / 255.\n",
" means.append(np.mean(total_imgs))\n",
" stds.append(np.std(total_imgs))\n",
"\n",
"\n",
"print('data mean is %f' % np.mean(means))\n",
"print('data std is %f' % np.std(stds))\n",
"print('total data size is %f' % count)\n",
"print('processing data end !')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"for i in range (1,41):\n",
" j=0\n",
" l=1\n",
" if(i<10):\n",
" os.mkdir('./testing/Patient_0'+str(i))\n",
" for k in os.listdir('./data/data_npy/Patient_0'+str(i)):\n",
" j=j+1\n",
" else:\n",
" os.mkdir('./testing/Patient_'+str(i))\n",
" for k in os.listdir('./data/data_npy/Patient_'+str(i)):\n",
" j=j+1\n",
" while(l<=(j/2)):\n",
" if(i<10):\n",
" lbl_array = np.load('./data/data_npy/Patient_0'+str(i)+'/Patient_0'+str(i)+'_'+str(l)+'_label.npy')\n",
" img = Image.fromarray(lbl_array)\n",
" matplotlib.image.imsave('./testing/Patient_0'+str(i)+'/Patient_0'+str(i)+'_'+str(l)+'_label.png', img)\n",
" else:\n",
" lbl_array = np.load('./data/data_npy/Patient_'+str(i)+'/Patient_'+str(i)+'_'+str(l)+'_label.npy')\n",
" img = Image.fromarray(lbl_array)\n",
" matplotlib.image.imsave('./testing/Patient_'+str(i)+'/Patient_'+str(i)+'_'+str(l)+'_label.png', img)\n",
" l=l+1\n",
"\n",
" \n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"for i in range(41,61):\n",
" j=0\n",
" l=1\n",
" os.mkdir('./testing_2d/Patient_'+str(i))\n",
" for k in os.listdir('./data_test/data_npy/Patient_'+str(i)+'.nii'):\n",
" j=j+1\n",
" while(l<=(j)):\n",
" img_array = np.load('./data_test/data_npy/Patient_'+str(i)+'.nii/Patient_'+str(i)+'.nii_'+str(l)+'_image.npy')\n",
" matplotlib.image.imsave('./testing_2d/Patient_'+str(i)+'/Patient_'+str(i)+'_'+str(l)+'_image.png', img_array)\n",
" l=l+1"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "gMZFGT4Qpr64"
},
"outputs": [],
"source": [
"%reload_ext autoreload\n",
"%autoreload 2\n",
"%matplotlib inline"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "5mXYcsKgpr7A"
},
"outputs": [],
"source": [
"from fastai.vision import *\n",
"from fastai.callbacks.hooks import *\n",
"from fastai.utils.mem import *\n",
"os.environ['CUDA_LAUNCH_BLOCKING'] = '1'"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
},
"colab_type": "code",
"id": "XcCEVOyHpr7G",
"outputId": "6119067e-b766-41b7-d650-dedf872a42d8",
"scrolled": true
},
"outputs": [],
"source": [
"cd train_2d/"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "xLTbzylqpr7a"
},
"outputs": [],
"source": [
"path_lbl = 'labels1/'\n",
"path_img = 'ct_images/'"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "8hALaOJzpr7m"
},
"source": [
"## Data"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 69
},
"colab_type": "code",
"id": "r-rnyaxSpr7o",
"outputId": "7cebd9b4-93f3-4ae8-a27f-55fc75e8a1a4"
},
"outputs": [],
"source": [
"fnames = get_image_files(path_img)\n",
"fnames[:3]"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 69
},
"colab_type": "code",
"id": "Uipy1i0Hpr73",
"outputId": "ca5ec985-3777-4fde-b1d8-e7072c0c4a19"
},
"outputs": [],
"source": [
"lbl_names = get_image_files(path_lbl)\n",
"lbl_names[:3]"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 323
},
"colab_type": "code",
"id": "aVJ8i78Dpr8A",
"outputId": "be54247f-31fd-4951-e28c-5f52a30b26de"
},
"outputs": [],
"source": [
"img_f = fnames[4]\n",
"img = open_image(img_f)\n",
"img.show(figsize=(5,5))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "_j57JSjipr8G"
},
"outputs": [],
"source": [
"get_y_fn = lambda x:path_lbl+f'{(x.stem)}_label{x.suffix}'"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"arr=np.load('Patient_01_100_image_label.npy')\n",
"x=plt.imshow(arr, cmap='Greys')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 323
},
"colab_type": "code",
"id": "vneNAIeipr8L",
"outputId": "49aba5d2-c715-4f53-a9fe-87a9aba4eb5b",
"scrolled": true
},
"outputs": [],
"source": [
"mask = open_mask(get_y_fn(img_f))\n",
"mask.show(figsize=(5,5), alpha=1)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 139
},
"colab_type": "code",
"id": "nVPrFLrupr8Q",
"outputId": "14a2f066-73ff-4298-8ab3-304b684f59f9",
"scrolled": true
},
"outputs": [],
"source": [
"src_size = np.array(mask.shape[1:])\n",
"src_size,mask.data"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
},
"colab_type": "code",
"id": "f3OXtvkSpr8b",
"outputId": "55f3ff6c-b31b-49c9-a269-7304c8bf9cdf"
},
"outputs": [],
"source": [
"codes = np.loadtxt('./codes.txt', dtype=str);codes#= np.delete(codes,4);codes"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "rrxWfznwpr8h"
},
"source": [
"## Datasets"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "lT9uhZeQpr8i"
},
"outputs": [],
"source": [
"size = src_size\n",
"\n",
"free = gpu_mem_get_free_no_cache()\n",
"# the max size of bs depends on the available GPU RAM\n",
"if free > 8200: bs=8\n",
"else: bs=4\n",
"print(f\"using bs={bs}, have {free}MB of GPU RAM free\")\n",
"bs=2"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
},
"colab_type": "code",
"id": "2wT3ahktpr8t",
"outputId": "e1d22796-9f8d-4904-de7c-b601cde0e6fe"
},
"outputs": [],
"source": [
"free = gpu_mem_get_free_no_cache()\n",
"free"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "xTeb9f-Npr84"
},
"outputs": [],
"source": [
"src = (SegmentationItemList.from_folder(path_img)\n",
" .split_by_rand_pct()\n",
" .label_from_func(get_y_fn, classes=codes))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "p6p9s3-Ypr88"
},
"outputs": [],
"source": [
"data = (src.transform(get_transforms(), size=size, tfm_y=True)\n",
" .databunch(bs=bs)\n",
" .normalize(imagenet_stats))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 513
},
"colab_type": "code",
"id": "S6yarrSKpr9E",
"outputId": "3c80296c-dfca-4e2c-fa32-d2b2746da3f8"
},
"outputs": [],
"source": [
"data.show_batch(2, figsize=(10,7))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 513
},
"colab_type": "code",
"id": "uhT57Xkgpr9M",
"outputId": "d9597c0e-7ba0-49af-826a-032fb469a77e"
},
"outputs": [],
"source": [
"data.show_batch(2, figsize=(10,7), ds_type=DatasetType.Valid)"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "zgo7kt5Bpr9R"
},
"source": [
"## Model"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "b5y3FjM9pr9R"
},
"outputs": [],
"source": [
"\n",
"def dice_coeff(input,target):\n",
" \"\"\"\n",
" input is a torch variable of size BatchxnclassesxHxW representing log probabilities for each class\n",
" target is a 1-hot representation of the groundtruth, shoud have same size as the input\n",
" \"\"\"\n",
" assert input.size() == target.size(), \"Input sizes must be equal.\"\n",
" assert input.dim() == 4, \"Input must be a 4D Tensor.\"\n",
" uniques=np.unique(target.numpy())\n",
" assert set(list(uniques))<=set([0,1]), \"target must only contain zeros and ones\"\n",
"\n",
" probs=F.softmax(input)\n",
" num=probs*target#b,c,h,w--p*g\n",
" num=torch.sum(num,dim=3)#b,c,h\n",
" num=torch.sum(num,dim=2)\n",
" \n",
"\n",
" den1=torch.sum(den1,dim=3)#b,c,h\n",
" den1=torch.sum(den1,dim=2)\n",
" \n",
"\n",
" den2=target*target#--g^2\n",
" den2=torch.sum(den2,dim=3)#b,c,h\n",
" den2=torch.sum(den2,dim=2)#b,c\n",
" \n",
"\n",
" dice=2*(num/(den1+den2))\n",
" dice_eso=dice[:,1:]#we ignore bg dice val, and take the fg\n",
"\n",
" dice_total=-1*torch.sum(dice_eso)/dice_eso.size(0)#divide by batch_sz\n",
"\n",
" return dice_total\n",
"def iou(target,prediction):\n",
" intersection = np.logical_and(target, prediction)\n",
" union = np.logical_or(target, prediction)\n",
" iou_score = np.sum(intersection) / np.sum(union)\n",
" return iou_score\n",
" \n",
"def soft_dice_loss(y_true, y_pred, epsilon=1e-6): \n",
" ''' \n",
" Soft dice loss calculation for arbitrary batch size, number of classes, and number of spatial dimensions.\n",
" Assumes the `channels_last` format.\n",
" \n",
" # Arguments\n",
" y_true: b x X x Y( x Z...) x c One hot encoding of ground truth\n",
" y_pred: b x X x Y( x Z...) x c Network output, must sum to 1 over c channel (such as after softmax) \n",
" epsilon: Used for numerical stability to avoid divide by zero errors\n",
" \n",
" # References\n",
" V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation \n",
" https://arxiv.org/abs/1606.04797\n",
" More details on Dice loss formulation \n",
" https://mediatum.ub.tum.de/doc/1395260/1395260.pdf (page 72)\n",
" \n",
" Adapted from https://github.com/Lasagne/Recipes/issues/99#issuecomment-347775022\n",
" '''\n",
" \n",
" # skip the batch and class axis for calculating Dice score\n",
" axes = tuple(range(1, len(y_pred.shape)-1)) \n",
" numerator = 2. * np.sum(y_pred * y_true, axes)\n",
" denominator = np.sum(np.square(y_pred) + np.square(y_true), axes)\n",
" \n",
" return 1 - np.mean(numerator / (denominator + epsilon)) # average over classes and batch\n",
"name2id = {v:k for k,v in enumerate(codes)}\n",
"void_code = name2id['Void']\n",
"\n",
"def acc(input, target):\n",
" target = target.squeeze(1)\n",
" mask = target != void_code\n",
" return (input.argmax(dim=1)[mask]==target[mask]).float().mean()\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "bfaw3BcKpr9V"
},
"outputs": [],
"source": [
"metrics=acc"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "n0EDeSXPpr9X"
},
"outputs": [],
"source": [
"wd=1e-2"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "-lMHeE3epr9Z"
},
"outputs": [],
"source": [
"learn = unet_learner(data, models.resnet18, metrics=metrics, wd=wd)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 1000
},
"colab_type": "code",
"id": "ib8EtiesiI3J",
"outputId": "8ae75823-8e78-40bc-d16a-34381763d4ba"
},
"outputs": [],
"source": [
"from torchsummary import summary\n",
"summary(learn.model, input_size=(3, 512, 512))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 1000
},
"colab_type": "code",
"id": "G7smN9QO9M0k",
"outputId": "0db355f0-588d-450e-a057-474c480456ba",
"scrolled": false
},
"outputs": [],
"source": [
"learn.model"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "lyYm11rFrbnk"
},
"outputs": [],
"source": [
"os.environ['CUDA_LAUNCH_BLOCKING'] = '1'"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 132
},
"colab_type": "code",
"id": "EzPVb8-9pr9i",
"outputId": "4fd99a5b-5147-4590-b22d-e1847a8c7102"
},
"outputs": [],
"source": [
"lr_find(learn)\n",
"learn.recorder.plot()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "tZztkldvpr9r"
},
"outputs": [],
"source": [
"lr=3e-3"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 391
},
"colab_type": "code",
"id": "y_zi_wBVpr9u",
"outputId": "e25606c6-cadf-474e-b99d-7f98a7da304a"
},
"outputs": [],
"source": [
"learn.fit_one_cycle(10, slice(lr), pct_start=0.9)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "SZdqGQrVpr9y"
},
"outputs": [],
"source": [
"learn.save('stage-1-imp')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "ZDIptQS_pr91"
},
"outputs": [],
"source": [
"learn.load('stage-1-imp');"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "FTEniKzepr95"
},
"outputs": [],
"source": [
"learn.show_results(rows=3, figsize=(8,9))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## POST-PROCESSING"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"img = open_image('./testing_2d/Patient_41/Patient_41_1_image.png')\n",
"img"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"im_g=learn.predict(img)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"scrolled": true
},
"outputs": [],
"source": [
"na1=(np.array(im_g[1][0]))\n",
"plt.imshow(arr)\n",
"plt.show()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"np.array(im_g[1][0])"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import matplotlib\n",
"from matplotlib import pyplot as plt"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"for i in range (41,61):\n",
" j=0\n",
" l=1\n",
" if(i<10):\n",
" os.mkdir('./testing/Patient_0'+str(i))\n",
" for k in os.listdir('./data/data_npy/Patient_0'+str(i)):\n",
" j=j+1\n",
" else:\n",
" os.mkdir('./testing/Patient_'+str(i))\n",
" for k in os.listdir('./testing_2d/Patient_'+str(i)):\n",
" j=j+1\n",
" while(l<=(j)):\n",
" if(i<10):\n",
" img = open_image('./testing_2d/Patient_0'+str(i)+'/Patient_0'+str(i)+'_'+str(l)+'_image.png')\n",
" im_g=learn.predict(img)\n",
" img = Image.fromarray(np.array(im_g[1][0]))\n",
" matplotlib.image.imsave('./testing/Patient_0'+str(i)+'/Patient_0'+str(i)+'_'+str(l)+'_image_label.png', np.array(im_g[1][0]))\n",
" else:\n",
" img = open_image('./testing_2d/Patient_'+str(i)+'/Patient_'+str(i)+'_'+str(l)+'_image.png')\n",
" im_g=learn.predict(img)\n",
" img = Image.fromarray(np.array(im_g[1][0]))\n",
" matplotlib.image.imsave('./testing/Patient_'+str(i)+'/Patient_'+str(i)+'_'+str(l)+'_image_label.png', np.array(im_g[1][0]))\n",
" l=l+1\n",
" "
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import nibabel as nib"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"scrolled": false
},
"outputs": [],
"source": [
"import nibabel as nib\n",
"for i in range(42,61):\n",
" j=0\n",
" l=1\n",
" img = open_image('./testing_2d/Patient_'+str(i)+'/Patient_'+str(i)+'_'+str(1)+'_image.png')\n",
" im_g=learn.predict(img)\n",
" na=np.array(im_g[1][0])\n",
" for k in os.listdir('./testing_2d/Patient_'+str(i)):\n",
" j=j+1\n",
" while(l<(j)):\n",
" img = open_image('./testing_2d/Patient_'+str(i)+'/Patient_'+str(i)+'_'+str(l+1)+'_image.png')\n",
" im_g=learn.predict(img)\n",
" a=np.array(im_g[1][0])\n",
" na=np.dstack((na,a))\n",
" l=l+1\n",
" ex=\"test/Patient_\"+str(i)+\".nii\"\n",
" img = nib.load(ex)\n",
" im1=np.array(img.get_affine())\n",
" new_image = nib.Nifti1Image(np.asarray(na,dtype=\"uint8\" ), affine=im1)\n",
" nib.save(new_image,'./testing3d/Patient_'+str(i)+'_GT.nii.gz')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"na.shape"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "IwakG2rrpr9_"
},
"outputs": [],
"source": [
"learn.unfreeze()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "uTsRJkcTpr-A"
},
"outputs": [],
"source": [
"lrs = slice(lr/400,lr/4)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "SAYQI7JDpr-C",
"outputId": "0dea8106-7ea3-4e30-e5d6-b6d4cf18c9f8",
"scrolled": true
},
"outputs": [],
"source": [
"learn.fit_one_cycle(12, lrs, pct_start=0.8)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "8mGVVW3ipr-H"
},
"outputs": [],
"source": [
"learn.save('stage-2');"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "BoNKO9vHpr-K"
},
"source": [
"## Go big"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "CUYh7uAFpr-L"
},
"source": [
"You may have to restart your kernel and come back to this stage if you run out of memory, and may also need to decrease `bs`."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
},
"colab_type": "code",
"id": "ntd0bQLxpr-M",
"outputId": "413ebf26-5f04-4e04-dae7-97a925d2e28e"
},
"outputs": [],
"source": [
"#learn.destroy() # uncomment once 1.0.46 is out\n",
"\n",
"size = src_size\n",
"\n",
"free = gpu_mem_get_free_no_cache()\n",
"# the max size of bs depends on the available GPU RAM\n",
"if free > 8200: bs=3\n",
"else: bs=1\n",
"print(f\"using bs={bs}, have {free}MB of GPU RAM free\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "1VnuPjdzpr-P"
},
"outputs": [],
"source": [
"data = (src.transform(get_transforms(), size=size, tfm_y=True)\n",
" .databunch(bs=bs)\n",
" .normalize(imagenet_stats))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 52
},
"colab_type": "code",
"id": "LqnUbPoTpr-T",
"outputId": "7c8d7f71-63e5-4a97-f3a0-4a2f87c0eed2",
"scrolled": true
},
"outputs": [],
"source": [
"learn = unet_learner(data, models.resnet50, metrics=metrics, wd=wd)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "ksf4YOD1pr-X"
},
"outputs": [],
"source": [
"learn.load('stage-2');"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 434
},
"colab_type": "code",
"id": "K8hZUjFzpr-Z",
"outputId": "7a03f3c3-968d-4270-c027-c1f534c79b0b"
},
"outputs": [],
"source": [
"lr_find(learn)\n",
"learn.recorder.plot()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "aRYdIM0lpr-b"
},
"outputs": [],
"source": [
"lr=1e-3"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "5u-cUGYtpr-c",
"outputId": "a66d77c2-4267-4b98-fe52-c54aec433552",
"scrolled": true
},
"outputs": [],
"source": [
"learn.fit_one_cycle(10, slice(lr), pct_start=0.8)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "TU38pO5gpr-f"
},
"outputs": [],
"source": [
"learn.save('stage-1-big')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "F2czKtoFpr-g"
},
"outputs": [],
"source": [
"learn.load('stage-1-big');"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "Oicfx00fpr-i"
},
"outputs": [],
"source": [
"learn.unfreeze()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "_6xQhCXApr-j"
},
"outputs": [],
"source": [
"lrs = slice(1e-6,lr/10)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "sxTN1UYSpr-l",
"outputId": "17f9b0c4-6a43-4182-a1f4-57b7ea235d4c"
},
"outputs": [],
"source": [
"learn.fit_one_cycle(10, lrs)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "tvm_yXCJpr-o"
},
"outputs": [],
"source": [
"learn.save('stage-2-big')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "7Kts9ls7pr-p"
},
"outputs": [],
"source": [
"learn.load('stage-2-big');"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "LektCRD2pr-y",
"outputId": "c73598d8-a91c-406d-a96e-022428a092cb"
},
"outputs": [],
"source": [
"learn.show_results(rows=3, figsize=(10,10))"
]
},
{
"cell_type": "markdown",
"metadata": {
"colab_type": "text",
"id": "dNCx0BaYpr-8"
},
"source": [
"## VISUALIZING Axial, Sagittal and Coronal"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"nii='Patient_41_GT.nii.gz'"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {},
"colab_type": "code",
"id": "ayG7b91Mpr-8",
"scrolled": true
},
"outputs": [],
"source": [
"import nilearn\n",
"from nilearn import plotting\n",
"plotting.plot_stat_map(nii)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"nii='GT.nii.gz'"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import nilearn\n",
"from nilearn import plotting\n",
"plotting.plot_stat_map(nii)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"accelerator": "GPU",
"colab": {
"collapsed_sections": [
"BoNKO9vHpr-K",
"dNCx0BaYpr-8"
],
"name": "SegThor.ipynb",
"provenance": [],
"version": "0.3.2"
},
"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": "3.7.3"
}
},
"nbformat": 4,
"nbformat_minor": 1
}