Apply transforms in PreProcessor (#2467)

* apply transforms in pre-processor * add augmentation arguments to datamodules * update expected config for adapter tests * fix buffer issue * update data notebooks * reduce num workers in MLFlow notebook * Revert "reduce num workers in MLFlow notebook" This reverts commit 151a179. * match resize between augmentations and model transforms * remove subset-specific transforms in preprocessor * move nested attr helper to utils * move transform retrieve function to transform utils * update efficientad transform validation * formatting * update preprocessor docstring * fix data notebook * fix logic in _update_augmentations * add unit tests for updating augmentations in data module * add unit tests for updating augmentations in data module * add test for collate method * copy transform before converting * update docstring * rename function
openvinotoolkit · Jan 6, 2025 · 235952e · 235952e
1 parent 8a82dc7
commit 235952e
Show file tree

Hide file tree

Showing 69 changed files with 762 additions and 728 deletions.
diff --git a/configs/model/cfa.yaml b/configs/model/cfa.yaml
@@ -8,9 +8,6 @@ model:
     num_hard_negative_features: 3
     radius: 1.0e-05
 
-metrics:
-  pixel: AUROC
-
 trainer:
   max_epochs: 30
   callbacks:

diff --git a/configs/model/cflow.yaml b/configs/model/cflow.yaml
@@ -15,10 +15,6 @@ model:
     permute_soft: false
     lr: 0.0001
 
-metrics:
-  pixel:
-    - AUROC
-
 trainer:
   max_epochs: 50
   callbacks:

diff --git a/configs/model/csflow.yaml b/configs/model/csflow.yaml
@@ -6,10 +6,6 @@ model:
     clamp: 3
     num_channels: 3
 
-metrics:
-  pixel:
-    - AUROC
-
 trainer:
   max_epochs: 240
   callbacks:

diff --git a/configs/model/draem.yaml b/configs/model/draem.yaml
@@ -6,10 +6,6 @@ model:
     sspcab_lambda: 0.1
     anomaly_source_path: null
 
-metrics:
-  pixel:
-    - AUROC
-
 trainer:
   max_epochs: 700
   callbacks:

diff --git a/configs/model/dsr.yaml b/configs/model/dsr.yaml
@@ -4,10 +4,6 @@ model:
     latent_anomaly_strength: 0.2
     upsampling_train_ratio: 0.7
 
-metrics:
-  pixel:
-    - AUROC
-
 # PL Trainer Args. Don't add extra parameter here.
 trainer:
   max_epochs: 700
diff --git a/configs/model/efficient_ad.yaml b/configs/model/efficient_ad.yaml
@@ -8,10 +8,6 @@ model:
     padding: false
     pad_maps: true
 
-metrics:
-  pixel:
-    - AUROC
-
 trainer:
   max_epochs: 1000
   max_steps: 70000
diff --git a/configs/model/fastflow.yaml b/configs/model/fastflow.yaml
@@ -7,10 +7,6 @@ model:
     conv3x3_only: false
     hidden_ratio: 1.0
 
-metrics:
-  pixel:
-    - AUROC
-
 trainer:
   max_epochs: 500
   callbacks:

diff --git a/configs/model/padim.yaml b/configs/model/padim.yaml
@@ -8,6 +8,3 @@ model:
     backbone: resnet18
     pre_trained: true
     n_features: null
-
-metrics:
-  pixel: AUROC
diff --git a/configs/model/reverse_distillation.yaml b/configs/model/reverse_distillation.yaml
@@ -9,10 +9,6 @@ model:
     anomaly_map_mode: ADD
     pre_trained: true
 
-metrics:
-  pixel:
-    - AUROC
-
 trainer:
   callbacks:
     - class_path: lightning.pytorch.callbacks.EarlyStopping

diff --git a/configs/model/stfpm.yaml b/configs/model/stfpm.yaml
@@ -7,10 +7,6 @@ model:
       - layer2
       - layer3
 
-metrics:
-  pixel:
-    - AUROC
-
 trainer:
   max_epochs: 100
   callbacks:

diff --git a/configs/model/uflow.yaml b/configs/model/uflow.yaml
@@ -7,10 +7,6 @@ model:
     affine_subnet_channels_ratio: 1.0
     backbone: mcait # official: mcait, other extractors tested: resnet18, wide_resnet50_2. Could use others...
 
-metrics:
-  pixel:
-    - AUROC
-
 # PL Trainer Args. Don't add extra parameter here.
 trainer:
   max_epochs: 200

diff --git a/notebooks/100_datamodules/101_btech.ipynb b/notebooks/100_datamodules/101_btech.ipynb
@@ -39,7 +39,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -61,18 +61,16 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 3,
    "metadata": {},
    "outputs": [],
    "source": [
     "# flake8: noqa\n",
     "import numpy as np\n",
     "from PIL import Image\n",
-    "from torchvision.transforms.v2 import Resize\n",
     "from torchvision.transforms.v2.functional import to_pil_image\n",
     "\n",
-    "from anomalib.data import BTech, BTechDataset\n",
-    "from anomalib import TaskType"
+    "from anomalib.data import BTech, BTechDataset"
    ]
   },
   {
@@ -99,7 +97,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 4,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -203,25 +201,6 @@
     "BTechDataset??"
    ]
   },
-  {
-   "attachments": {},
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "We can add some transforms that will be applied to the images using torchvision. Let's add a transform that resizes the \n",
-    "input image to 256x256 pixels."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "image_size = (256, 256)\n",
-    "transform = Resize(image_size, antialias=True)"
-   ]
-  },
   {
    "attachments": {},
    "cell_type": "markdown",
@@ -240,7 +219,6 @@
     "btech_dataset_train = BTechDataset(\n",
     "    root=dataset_root,\n",
     "    category=\"01\",\n",
-    "    transform=transform,\n",
     "    split=\"train\",\n",
     ")\n",
     "print(len(btech_dataset_train))\n",
@@ -268,7 +246,6 @@
     "btech_dataset_test = BTechDataset(\n",
     "    root=dataset_root,\n",
     "    category=\"01\",\n",
-    "    transform=transform,\n",
     "    split=\"test\",\n",
     ")\n",
     "print(len(btech_dataset_test))\n",

diff --git a/notebooks/100_datamodules/102_mvtec.ipynb b/notebooks/100_datamodules/102_mvtec.ipynb
@@ -23,14 +23,13 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
     "# flake8: noqa\n",
     "import numpy as np\n",
     "from PIL import Image\n",
-    "from torchvision.transforms.v2 import Resize\n",
     "from torchvision.transforms.v2.functional import to_pil_image\n",
     "\n",
     "from anomalib.data import MVTec, MVTecDataset"
@@ -48,7 +47,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 3,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -76,7 +75,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 4,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -180,25 +179,6 @@
     "MVTecDataset??"
    ]
   },
-  {
-   "attachments": {},
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "We can add some transforms that will be applied to the images using torchvision. Let's add a transform that resizes the \n",
-    "input image to 256x256 pixels."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "image_size = (256, 256)\n",
-    "transform = Resize(image_size, antialias=True)"
-   ]
-  },
   {
    "attachments": {},
    "cell_type": "markdown",
@@ -217,7 +197,6 @@
     "mvtec_dataset_train = MVTecDataset(\n",
     "    root=dataset_root,\n",
     "    category=\"bottle\",\n",
-    "    transform=transform,\n",
     "    split=\"train\",\n",
     ")\n",
     "print(len(mvtec_dataset_train))\n",
@@ -245,7 +224,6 @@
     "mvtec_dataset_test = MVTecDataset(\n",
     "    root=dataset_root,\n",
     "    category=\"bottle\",\n",
-    "    transform=transform,\n",
     "    split=\"test\",\n",
     ")\n",
     "print(len(mvtec_dataset_test))\n",

diff --git a/notebooks/100_datamodules/103_folder.ipynb b/notebooks/100_datamodules/103_folder.ipynb
@@ -33,7 +33,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -63,14 +63,13 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 3,
    "metadata": {},
    "outputs": [],
    "source": [
     "# flake8: noqa\n",
     "import numpy as np\n",
     "from PIL import Image\n",
-    "from torchvision.transforms.v2 import Resize\n",
     "from torchvision.transforms.v2.functional import to_pil_image\n",
     "\n",
     "from anomalib.data import Folder, FolderDataset"
@@ -173,25 +172,6 @@
     "FolderDataset??"
    ]
   },
-  {
-   "attachments": {},
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "We can add some transforms that will be applied to the images using torchvision. Let's add a transform that resizes the \n",
-    "input image to 256x256 pixels."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "image_size = (256, 256)\n",
-    "transform = Resize(image_size, antialias=True)"
-   ]
-  },
   {
    "attachments": {},
    "cell_type": "markdown",
@@ -211,7 +191,6 @@
     "    normal_dir=dataset_root / \"good\",\n",
     "    abnormal_dir=dataset_root / \"crack\",\n",
     "    split=\"train\",\n",
-    "    transform=transform,\n",
     ")\n",
     "print(len(folder_dataset_train))\n",
     "sample = folder_dataset_train[0]\n",
@@ -241,7 +220,6 @@
     "    normal_dir=dataset_root / \"good\",\n",
     "    abnormal_dir=dataset_root / \"crack\",\n",
     "    split=\"test\",\n",
-    "    transform=transform,\n",
     ")\n",
     "print(len(folder_dataset_test))\n",
     "sample = folder_dataset_test[0]\n",
@@ -270,7 +248,6 @@
     "    normal_dir=dataset_root / \"good\",\n",
     "    abnormal_dir=dataset_root / \"crack\",\n",
     "    split=\"train\",\n",
-    "    transform=transform,\n",
     "    mask_dir=dataset_root / \"mask\" / \"crack\",\n",
     ")\n",
     "print(len(folder_dataset_segmentation_train))\n",
@@ -290,7 +267,6 @@
     "    normal_dir=dataset_root / \"good\",\n",
     "    abnormal_dir=dataset_root / \"crack\",\n",
     "    split=\"test\",\n",
-    "    transform=transform,\n",
     "    mask_dir=dataset_root / \"mask\" / \"crack\",\n",
     ")\n",
     "print(len(folder_dataset_segmentation_test))\n",