Refactor Visual Prompting OV model to use MAPI (#3638)

* Use SAM OV models from MAPI

* Disable throughput for vpt becase of sync infer

* Update vpt exporter

* Replace infer in ZSL with MAPI

* Replace infer in ZSL with MAPI

* Fix input name for ZSL wrapper

* Fix merge artifacts

* Cleanup

* Adapt ZSL optimize to MAPI

* Del unused code

* Refactor infer in zsl

* Update SAM and ZSL MAPI usage

* Adapt to changes in MAPI

* Fix import sorting

* Bump MAPI version

* Del outdated unit tests

* Fix tiling utest

* Fix vpt onnx export

* Fix visual prompting unit tests

* Fix linters

* Fix setup script

* Add todo

src/otx/algo/visual_prompting/__init__.py

@@ -1,10 +1,9 @@
-# Copyright (C) 2023 Intel Corporation
+# Copyright (C) 2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 #
 """Module for OTX visual prompting models."""
 
 from . import backbones, decoders, encoders
-from .openvino_models import VisualPromptingDecoder, VisualPromptingImageEncoder
 from .segment_anything import OTXSegmentAnything, SegmentAnything
 from .zero_shot_segment_anything import OTXZeroShotSegmentAnything, ZeroShotSegmentAnything
 
@@ -16,6 +15,4 @@
     "SegmentAnything",
     "OTXZeroShotSegmentAnything",
     "ZeroShotSegmentAnything",
-    "VisualPromptingImageEncoder",
-    "VisualPromptingDecoder",
 ]

src/otx/core/exporter/visual_prompting.py

@@ -7,6 +7,7 @@
 
 import logging as log
 import tempfile
+from functools import partial
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Literal
 
@@ -61,22 +62,23 @@ def export(  # type: ignore[override]
                 log.warning(msg)
             fn = self.to_openvino
         elif export_format == OTXExportFormatType.ONNX:
-            fn = self.to_onnx
+            fn = partial(self.to_onnx, embed_metadata=True)
         else:
             msg = f"Unsupported export format: {export_format}"
             raise ValueError(msg)
 
         return {  # type: ignore[return-value]
-            module: fn(models[module], output_dir, f"{base_model_name}_{module}", precision)
+            module: fn(models[module], output_dir, f"{base_model_name}_{module}", precision, model_type=f"sam_{module}")
             for module in ["image_encoder", "decoder"]
         }
 
     def to_openvino(
         self,
-        model: OTXModel,
+        model: OTXModel | torch.nn.Module,
         output_dir: Path,
         base_model_name: str = "exported_model",
         precision: OTXPrecisionType = OTXPrecisionType.FP32,
+        model_type: str = "sam",
     ) -> Path:
         """Export to OpenVINO Intermediate Representation format.
 
@@ -93,12 +95,17 @@ def to_openvino(
                 tmp_dir,
                 base_model_name,
                 OTXPrecisionType.FP32,
-                False,
+                embed_metadata=False,
             )
             exported_model = openvino.convert_model(tmp_dir / (base_model_name + ".onnx"))
 
         exported_model = self._postprocess_openvino_model(exported_model)
 
+        if self.metadata is not None:
+            export_metadata = self._extend_model_metadata(self.metadata)
+            export_metadata[("model_info", "model_type")] = model_type
+            exported_model = self._embed_openvino_ir_metadata(exported_model, export_metadata)
+
         save_path = output_dir / (base_model_name + ".xml")
         openvino.save_model(exported_model, save_path, compress_to_fp16=(precision == OTXPrecisionType.FP16))
         log.info("Converting to OpenVINO is done.")
@@ -107,11 +114,12 @@ def to_openvino(
 
     def to_onnx(
         self,
-        model: OTXModel,
+        model: OTXModel | torch.nn.Module,
         output_dir: Path,
         base_model_name: str = "exported_model",
         precision: OTXPrecisionType = OTXPrecisionType.FP32,
         embed_metadata: bool = True,
+        model_type: str = "sam",
     ) -> Path:
         """Export the given PyTorch model to ONNX format and save it to the specified output directory.
 
@@ -136,7 +144,12 @@ def to_onnx(
         )
 
         onnx_model = onnx.load(save_path)
-        onnx_model = self._postprocess_onnx_model(onnx_model, embed_metadata, precision)
+        onnx_model = self._postprocess_onnx_model(onnx_model, False, precision)
+
+        if self.metadata is not None and embed_metadata:
+            export_metadata = self._extend_model_metadata(self.metadata)
+            export_metadata[("model_info", "model_type")] = model_type
+            onnx_model = self._embed_onnx_metadata(onnx_model, export_metadata)
 
         onnx.save(onnx_model, save_path)
         log.info("Converting to ONNX is done.")