Skip to content

Training segmentation example notebook fixes #731

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
raffeyC wants to merge 7 commits into
base: master
Choose a base branch
from
Open

Training segmentation example notebook fixes #731

Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
a6bc422
Add missing tempfile import
raffeyC Aug 29, 2024
40842a7
Change data_dir to work with docker by default
raffeyC Aug 29, 2024
0b33cac
Add missing radius parameter
raffeyC Aug 29, 2024
6aad666
Fix typo for histogram_normalization
raffeyC Aug 29, 2024
f245a59
Prepare train data for the pretrained model
raffeyC Aug 29, 2024
3f8f8a7
Change X_test to X_crop and markdown changes
raffeyC Aug 29, 2024
2e790cf
Merge branch 'master' into train_segmentation_fixes
rossbar Aug 30, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
101 changes: 93 additions & 8 deletions notebooks/training/panopticnets/Nuclear Segmentation - DeepWatershed.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@
"import numpy as np\n",
"from skimage.feature import peak_local_max\n",
"import tensorflow as tf\n",
"import tempfile\n",
"\n",
"from deepcell.applications import NuclearSegmentation\n",
"from deepcell.image_generators import CroppingDataGenerator\n",
Expand All @@ -54,7 +55,7 @@
"metadata": {},
"outputs": [],
"source": [
"data_dir = '/notebooks/data'\n",
"data_dir = '/data'\n",
"model_path = 'NuclearSegmentation'\n",
"metrics_path = 'metrics.yaml'\n",
"train_log = 'train_log.csv'"
Expand Down Expand Up @@ -146,6 +147,7 @@
"outputs": [],
"source": [
"# Post processing parameters\n",
"radius = 10\n",
"maxima_threshold = 0.1\n",
"interior_threshold = 0.01\n",
"exclude_border = False\n",
Expand Down Expand Up @@ -187,6 +189,7 @@
"X_train = histogram_normalization(X_train)\n",
"X_val = histogram_normalization(X_val)\n",
"\n",
"\n",
"# use augmentation for training but not validation\n",
"datagen = CroppingDataGenerator(\n",
" rotation_range=180,\n",
Expand Down Expand Up @@ -452,15 +455,72 @@
"## Predict on test data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Prepare data by using validation data generator"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"X_test = histograph_normalization(X_test)\n",
"X_test = histogram_normalization(X_test)\n",
"\n",
"test_images = prediction_model.predict(X_test)"
"# Generator used to crop data and transform y\n",
"test_data = datagen_val.flow(\n",
" {'X': X_test, 'y': y_test},\n",
" seed=seed,\n",
" min_objects=min_objects,\n",
" transforms=transforms,\n",
" transforms_kwargs=transforms_kwargs,\n",
" batch_size=batch_size,\n",
")\n",
"\n",
"# Generator used to crop y without transform\n",
"test_data_y = datagen_val.flow(\n",
" {'X': y_test, 'y': y_test},\n",
" seed=seed,\n",
" min_objects=min_objects,\n",
" transforms=[],\n",
" transforms_kwargs={},\n",
" batch_size=batch_size,\n",
")\n",
"\n",
"X_crop, y_crop, y_crop_t = None, None, None\n",
"for i, j in test_data:\n",
" \n",
" X_crop = np.concatenate((X_crop, i), axis=0) if X_crop is not None else i\n",
" # select needed transform as y_crop_t\n",
" y_crop_t = np.concatenate((y_crop_t, j[2]), axis=0) if y_crop_t is not None else j[2]\n",
"\n",
" if len(X_crop)>=len(X_test):\n",
" print(X_crop.shape)\n",
" \n",
" break\n",
"\n",
"for i, _ in test_data_y:\n",
" if y_crop is None:\n",
" y_crop = i\n",
"\n",
" elif len(y_crop)>=len(y_test):\n",
" print(y_crop.shape)\n",
" break\n",
" \n",
" else:\n",
" y_crop = np.concatenate((y_crop, i), axis=0)\n",
"\n",
"\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Predict and visualize"
]
},
{
Expand All @@ -471,7 +531,8 @@
},
"outputs": [],
"source": [
"index = np.random.choice(X_test.shape[0])\n",
"test_images = prediction_model.predict(X_crop)\n",
"index = np.random.choice(X_crop.shape[0])\n",
"print(index)\n",
"\n",
"fig, axes = plt.subplots(1, 4, figsize=(20, 20))\n",
Expand All @@ -496,7 +557,7 @@
")\n",
"\n",
"# raw image with centroid\n",
"axes[0].imshow(X_test[index, ..., 0])\n",
"axes[0].imshow(X_crop[index, ..., 0])\n",
"axes[0].scatter(coords[..., 1], coords[..., 0],\n",
" color='r', marker='.', s=10)\n",
"\n",
Expand Down Expand Up @@ -524,7 +585,7 @@
},
"outputs": [],
"source": [
"outputs = model.predict(X_test)\n",
"outputs = model.predict(X_crop)\n",
"\n",
"y_pred = []\n",
"\n",
Expand All @@ -542,12 +603,36 @@
" y_pred.append(mask[0])\n",
"\n",
"y_pred = np.stack(y_pred, axis=0)\n",
"y_pred = np.expand_dims(y_pred, axis=-1)\n",
"y_true = y_test.copy()\n",
"y_true = y_crop_t[:, :, :, 0].copy().astype(int)\n",
"y_true = np.expand_dims(y_true, axis=-1)\n",
"\n",
"m = Metrics('DeepWatershed', seg=False)\n",
"m.calc_object_stats(y_true, y_pred)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Visual check"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"for _ in range(5):\n",
" index = np.random.choice(X_crop.shape[0])\n",
" print(index)\n",
" fig, axes = plt.subplots(1, 3, figsize=(20, 20))\n",
" \n",
" axes[0].imshow(X_crop[index, ..., 0], cmap='jet')\n",
" axes[1].imshow(y_true[index, ..., 0], cmap='jet')\n",
" axes[2].imshow(y_pred[index, ..., 0], cmap='jet')\n",
" plt.show()"
]
}
],
"metadata": {
Expand Down