-
+
**A library for benchmarking, developing and deploying deep learning anomaly detection algorithms**
@@ -8,23 +8,40 @@
[Key Features](#key-features) •
[Docs](https://anomalib.readthedocs.io/en/latest/) •
-[Notebooks](notebooks) •
+[Notebooks](examples/notebooks) •
[License](LICENSE)
-[]()
-[]()
-[]()
+[]()
+[]()
+[]()
+[]()
[](https://github.com/openvinotoolkit/anomalib/actions/workflows/pre_merge.yml)
-[](https://anomalib.readthedocs.io/en/latest/?badge=latest)
[](https://codecov.io/gh/openvinotoolkit/anomalib)
[](https://pepy.tech/project/anomalib)
-[](https://discord.com/channels/1230798452577800237)
+
+[](https://anomalib.readthedocs.io/en/latest/?badge=latest)
+[](https://gurubase.io/g/anomalib)
![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174126337-b344bbdc-6343-4d85-93e8-0cb1bf39a4e3.png\")
\n",
- "\n",
- "| Class | Yellow cube | Red cube | Green cube | Inferencing using Anomalib |\n",
- "| -------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- |\n",
- "| Normal | ![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174083561-38eec918-efc2-4ceb-99b1-bbb4c91396b2.jpg\")
| ![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174083638-85ff889c-6222-4428-9c7d-9ad62bd15afe.jpg\")
| ![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174083707-364177d4-373b-4891-96ce-3e5ea923e440.jpg\")
| ![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174129305-03d9b71c-dfd9-492f-b42e-01c5c24171cc.jpg\")
|\n",
- "| Abnormal | ![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174083805-df0a0b03-58c7-4ba8-af50-fd94d3a13e58.jpg\")
| ![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174083873-22699523-22b4-4a55-a3da-6520095af8af.jpg\")
| ![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174083944-38d5a6f4-f647-455b-ba4e-69482dfa3562.jpg\")
| ![\"drawing\"](\"https://user-images.githubusercontent.com/10940214/174129253-f7a567d0-84f7-4050-8065-f00ba8bb973d.jpg\")
|\n",
- "\n",
- "Using Anomalib we are expecting to see this result.\n"
- ]
- },
- {
- "attachments": {},
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "# Installing Anomalib\n",
- "\n",
- "To install anomalib with the required dependencies, please follow the steps under `Install from source` [on GitHub](https://github.com/openvinotoolkit/anomalib?tab=readme-ov-file#-installation)."
- ]
- },
- {
- "attachments": {},
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "## Imports\n"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 1,
- "metadata": {
- "ExecuteTime": {
- "end_time": "2024-01-12T16:55:35.855912923Z",
- "start_time": "2024-01-12T16:55:30.140865729Z"
- }
- },
- "outputs": [],
- "source": [
- "\"\"\"501a_training_a_model_with_cubes_from_a_robotic_arm.ipynb.\"\"\"\n",
- "\n",
- "from pathlib import Path\n",
- "\n",
- "from anomalib.data.utils import read_image\n",
- "from anomalib.deploy import OpenVINOInferencer"
- ]
- },
- {
- "attachments": {},
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "## Download dataset and Robot API/Driver\n",
- "\n",
- "We should prepare the folder to save the dataset and the Dobot API and drivers. To download the dataset and the Dobot API and drivers we will use anomalib's `download_and_extract` utility function.\n"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 2,
- "metadata": {
- "ExecuteTime": {
- "end_time": "2024-01-12T16:55:40.293464182Z",
- "start_time": "2024-01-12T16:55:35.850186281Z"
- }
- },
- "outputs": [
- {
- "name": "stderr",
- "output_type": "stream",
- "text": [
- "cubes.zip: 6.99MB [00:01, 5.86MB/s] \n",
- "dobot_api.zip: 3.69MB [00:00, 5.43MB/s] \n"
- ]
- }
- ],
- "source": [
- "from anomalib.data.utils import DownloadInfo, download_and_extract\n",
- "\n",
- "dataset_download_info = DownloadInfo(\n",
- " name=\"cubes.zip\",\n",
- " url=\"https://github.com/openvinotoolkit/anomalib/releases/download/dobot/cubes.zip\",\n",
- " hashsum=\"182ce0a48dabf452bf9a6aeb83132466088e30ed7a5c35d7d3a10a9fc11daac4\",\n",
- ")\n",
- "api_download_info = DownloadInfo(\n",
- " name=\"dobot_api.zip\",\n",
- " url=\"https://github.com/openvinotoolkit/anomalib/releases/download/dobot/dobot_api.zip\",\n",
- " hashsum=\"eb79bb9c6346be1628a0fe5e1196420dcc4e122ab1aa0d5abbc82f63236f0527\",\n",
- ")\n",
- "download_and_extract(root=Path.cwd(), info=dataset_download_info)\n",
- "download_and_extract(root=Path.cwd(), info=api_download_info)"
- ]
- },
- {
- "attachments": {},
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "### Dataset: Cubes\n",
- "\n",
- "Prepare your own dataset for normal and defect pieces.\n"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 3,
- "metadata": {
- "ExecuteTime": {
- "end_time": "2024-01-12T16:55:40.725983993Z",
- "start_time": "2024-01-12T16:55:40.274675101Z"
- }
- },
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "dict_keys(['image_path', 'label', 'image'])\n"
- ]
- }
- ],
- "source": [
- "from anomalib import TaskType\n",
- "from anomalib.data import Folder\n",
- "\n",
- "datamodule = Folder(\n",
- " name=\"cubes\",\n",
- " root=Path.cwd() / \"cubes\",\n",
- " normal_dir=\"normal\",\n",
- " abnormal_dir=\"abnormal\",\n",
- " normal_split_ratio=0.2,\n",
- " image_size=(256, 256),\n",
- " train_batch_size=32,\n",
- " eval_batch_size=32,\n",
- " task=TaskType.CLASSIFICATION,\n",
- ")\n",
- "datamodule.setup()\n",
- "\n",
- "i, data = next(enumerate(datamodule.val_dataloader()))\n",
- "print(data.keys())"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 4,
- "metadata": {
- "ExecuteTime": {
- "end_time": "2024-01-12T16:55:40.734861023Z",
- "start_time": "2024-01-12T16:55:40.727834331Z"
- }
- },
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "torch.Size([32, 3, 256, 256])\n"
- ]
- }
- ],
- "source": [
- "# Check image size\n",
- "print(data[\"image\"].shape)"
- ]
- },
- {
- "attachments": {},
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "## Model\n",
- "\n",
- "`anomalib` supports a wide range of unsupervised anomaly detection models. The table in this [link](https://anomalib.readthedocs.io/en/latest/markdown/guides/reference/models/image/index.html) shows the list of models currently supported by `anomalib` library.\n"
- ]
- },
- {
- "attachments": {},
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "### Prepare the Model\n",
- "\n",
- "We will use Padim model for this use case, which could be imported from `anomalib.models`.\n"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 5,
- "metadata": {
- "ExecuteTime": {
- "end_time": "2024-01-12T16:55:41.184026691Z",
- "start_time": "2024-01-12T16:55:40.731669374Z"
- }
- },
- "outputs": [],
- "source": [
- "from anomalib.models import Padim\n",
- "\n",
- "model = Padim(\n",
- " backbone=\"resnet18\",\n",
- " layers=[\"layer1\", \"layer2\", \"layer3\"],\n",
- ")"
- ]
- },
- {
- "attachments": {},
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "## Training\n",
- "\n",
- "Now that we set up the datamodule and model, we could now train the model.\n",
- "\n",
- "The final component to train the model is `Engine` object, which handles train/test/predict/export pipeline. Let's create the engine object to train the model.\n"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 6,
- "metadata": {
- "ExecuteTime": {
- "end_time": "2024-01-12T16:55:45.425314142Z",
- "start_time": "2024-01-12T16:55:41.180954949Z"
- },
- "scrolled": true
- },
- "outputs": [
- {
- "name": "stderr",
- "output_type": "stream",
- "text": [
- "/home/djameln/miniconda3/envs/anomalibv1source/lib/python3.10/site-packages/torchmetrics/utilities/prints.py:36: UserWarning: Metric `PrecisionRecallCurve` will save all targets and predictions in buffer. For large datasets this may lead to large memory footprint.\n",
- " warnings.warn(*args, **kwargs)\n",
- "GPU available: True (cuda), used: True\n",
- "TPU available: False, using: 0 TPU cores\n",
- "IPU available: False, using: 0 IPUs\n",
- "HPU available: False, using: 0 HPUs\n",
- "`Trainer(val_check_interval=1.0)` was configured so validation will run at the end of the training epoch..\n",
- "You are using a CUDA device ('NVIDIA GeForce RTX 3090') that has Tensor Cores. To properly utilize them, you should set `torch.set_float32_matmul_precision('medium' | 'high')` which will trade-off precision for performance. For more details, read https://pytorch.org/docs/stable/generated/torch.set_float32_matmul_precision.html#torch.set_float32_matmul_precision\n",
- "/home/djameln/miniconda3/envs/anomalibv1source/lib/python3.10/site-packages/torchmetrics/utilities/prints.py:36: UserWarning: Metric `ROC` will save all targets and predictions in buffer. For large datasets this may lead to large memory footprint.\n",
- " warnings.warn(*args, **kwargs)\n",
- "LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0,1]\n",
- "/home/djameln/miniconda3/envs/anomalibv1source/lib/python3.10/site-packages/lightning/pytorch/core/optimizer.py:180: `LightningModule.configure_optimizers` returned `None`, this fit will run with no optimizer\n",
- "\n",
- " | Name | Type | Params\n",
- "-------------------------------------------------------------------\n",
- "0 | model | PadimModel | 2.8 M \n",
- "1 | _transform | Compose | 0 \n",
- "2 | normalization_metrics | MinMax | 0 \n",
- "3 | image_threshold | F1AdaptiveThreshold | 0 \n",
- "4 | pixel_threshold | F1AdaptiveThreshold | 0 \n",
- "5 | image_metrics | AnomalibMetricCollection | 0 \n",
- "6 | pixel_metrics | AnomalibMetricCollection | 0 \n",
- "-------------------------------------------------------------------\n",
- "2.8 M Trainable params\n",
- "0 Non-trainable params\n",
- "2.8 M Total params\n",
- "11.131 Total estimated model params size (MB)\n"
- ]
- },
- {
- "data": {
- "application/vnd.jupyter.widget-view+json": {
- "model_id": "cd7082c9791c4745a28c995a0f50abc8",
- "version_major": 2,
- "version_minor": 0
- },
- "text/plain": [
- "Training: | | 0/? [00:00┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓\n",
- "┃