added sftpm model fine-tuned on capsule category of the MVTec dataset

.gitattributes

@@ -32,3 +32,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+openvino_model.tar filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,51 @@
 ---
 license: apache-2.0
 ---
+[SFTPM](https://github.com/openvinotoolkit/anomalib/tree/main/anomalib/models/stfpm) model from [Anomalib](https://github.com/openvinotoolkit/anomalib) fine-tuned for capsule category of the [MVTec dataset](https://www.mvtec.com/company/research/datasets/mvtec-ad). Checkpoint trained using the following [notebook](https://github.com/openvinotoolkit/anomalib/blob/main/notebooks/000_getting_started/001_getting_started.ipynb).
+
+```
+──────────────────────────────────────────────────
+       Test metric             DataLoader 0
+──────────────────────────────────────────────────
+       image_AUROC          0.8436378240585327
+      image_F1Score         0.9356223344802856
+       pixel_AUROC          0.9719913601875305
+      pixel_F1Score         0.41566985845565796
+──────────────────────────────────────────────────
+```
+
+The main intent is to use it in samples and demos for model optimization. Here is the advantages:
+- MVTec dataset can automatically downloaded and is quite small.
+- The model from the anomaly detection domain such as SFTPM is sensitive to the optimization methods to allows demonstrate methods with accuracy controll.
+
+Here is the code to test the checkpoint:
+
+```python
+from pytorch_lightning import Trainer
+from anomalib.config import get_configurable_parameters
+from anomalib.data import get_datamodule
+from anomalib.models import get_model
+from anomalib.utils.callbacks import LoadModelCallback, get_callbacks
+
+CHECKPOINT_URL = 'https://huggingface.co/alexsu52/sftpm_mvtec_capsule/resolve/main/pytorch_model.bin'
+CHECKPOINT_PATH = '~/pytorch_model.bin'
+
+#Download CHECKPOINT_URL to CHECKPOINT_PATH
+
+config = get_configurable_parameters(config_path="./anomalib/models/sftpm/config.yaml")
+config["dataset"]["path"] = <path_to_dataset>
+config['dataset']['category'] = 'capsule'
+
+datamodule = get_datamodule(config)
+datamodule.setup()  # Downloads the dataset if it's not in the specified `root` directory
+datamodule.prepare_data()  # Create train/val/test/prediction sets.
+
+model = get_model(config)
+
+callbacks = get_callbacks(config)
+load_model_callback = LoadModelCallback(weights_path=CHECKPOINT_PATH)
+callbacks.insert(0, load_model_callback)
+
+trainer = Trainer(**config.trainer, callbacks=callbacks)
+trainer.test(model=model, datamodule=datamodule)
+```

openvino_model.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8420e799269d0b29244fdb32a5b670c3fb63fa8bf56d16547b02774793144b77
+size 44748800

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f3dff45bdddcb077cf2fa5eb7499b3746891f53059aa90bb59da068e62f93c8
+size 33536819

sftpm_capsule.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1dce6ebf6e78428a4d38ff0b78423cbd1e1ec973e1e07954d74fe64435d2428
+size 22522197