Add OpenVINO metadata to export (#7947)

* Write .yaml file when exporting model to openvino

Write .yaml file automatically when exporting model to openvino to be used during inference

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update export.py

* Update export.py

* Load metadata on inference

* Update common.py

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Glenn Jocher <[email protected]>

export.py

@@ -54,6 +54,7 @@ from pathlib import Path
 
 import pandas as pd
 import torch
+import yaml
 from torch.utils.mobile_optimizer import optimize_for_mobile
 
 FILE = Path(__file__).resolve()
@@ -168,7 +169,7 @@ def export_onnx(model, im, file, opset, train, dynamic, simplify, prefix=colorst
         LOGGER.info(f'{prefix} export failure: {e}')
 
 
-def export_openvino(file, half, prefix=colorstr('OpenVINO:')):
+def export_openvino(model, file, half, prefix=colorstr('OpenVINO:')):
     # YOLOv5 OpenVINO export
     try:
         check_requirements(('openvino-dev',))  # requires openvino-dev: https://pypi.org/project/openvino-dev/
@@ -178,7 +179,9 @@ def export_openvino(file, half, prefix=colorstr('OpenVINO:')):
         f = str(file).replace('.pt', f'_openvino_model{os.sep}')
 
         cmd = f"mo --input_model {file.with_suffix('.onnx')} --output_dir {f} --data_type {'FP16' if half else 'FP32'}"
-        subprocess.check_output(cmd.split())
+        subprocess.check_output(cmd.split())  # export
+        with open(Path(f) / 'meta.yaml', 'w') as g:
+            yaml.dump({'stride': int(max(model.stride)), 'names': model.names}, g)  # add metadata.yaml
 
         LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
         return f
@@ -520,7 +523,7 @@ def run(
     if onnx or xml:  # OpenVINO requires ONNX
         f[2] = export_onnx(model, im, file, opset, train, dynamic, simplify)
     if xml:  # OpenVINO
-        f[3] = export_openvino(file, half)
+        f[3] = export_openvino(model, file, half)
     if coreml:
         _, f[4] = export_coreml(model, im, file, int8, half)
 

models/common.py

@@ -326,9 +326,6 @@ class DetectMultiBackend(nn.Module):
         stride, names = 32, [f'class{i}' for i in range(1000)]  # assign defaults
         w = attempt_download(w)  # download if not local
         fp16 &= (pt or jit or onnx or engine) and device.type != 'cpu'  # FP16
-        if data:  # data.yaml path (optional)
-            with open(data, errors='ignore') as f:
-                names = yaml.safe_load(f)['names']  # class names
 
         if pt:  # PyTorch
             model = attempt_load(weights if isinstance(weights, list) else w, map_location=device)
@@ -367,7 +364,8 @@ class DetectMultiBackend(nn.Module):
                 w = next(Path(w).glob('*.xml'))  # get *.xml file from *_openvino_model dir
             network = ie.read_model(model=w, weights=Path(w).with_suffix('.bin'))
             executable_network = ie.compile_model(model=network, device_name="CPU")
-            self.output_layer = next(iter(executable_network.outputs))
+            output_layer = next(iter(executable_network.outputs))
+            self._load_metadata(w.parent / 'meta.yaml')  # load metadata
         elif engine:  # TensorRT
             LOGGER.info(f'Loading {w} for TensorRT inference...')
             import tensorrt as trt  # https://developer.nvidia.com/nvidia-tensorrt-download
@@ -433,7 +431,11 @@ class DetectMultiBackend(nn.Module):
                 output_details = interpreter.get_output_details()  # outputs
             elif tfjs:
                 raise Exception('ERROR: YOLOv5 TF.js inference is not supported')
+
         self.__dict__.update(locals())  # assign all variables to self
+        if not hasattr(self, 'names') and data:  # assign class names (optional)
+            with open(data, errors='ignore') as f:
+                names = yaml.safe_load(f)['names']
 
     def forward(self, im, augment=False, visualize=False, val=False):
         # YOLOv5 MultiBackend inference
@@ -493,13 +495,20 @@ class DetectMultiBackend(nn.Module):
             y = torch.tensor(y, device=self.device)
         return (y, []) if val else y
 
+    def _load_metadata(self, f='path/to/meta.yaml'):
+        # Load metadata from meta.yaml if it exists
+        if Path(f).is_file():
+            with open(f, errors='ignore') as f:
+                for k, v in yaml.safe_load(f).items():
+                    setattr(self, k, v)  # assign stride, names
+
     def warmup(self, imgsz=(1, 3, 640, 640)):
         # Warmup model by running inference once
-        if any((self.pt, self.jit, self.onnx, self.engine, self.saved_model, self.pb)):  # warmup types
-            if self.device.type != 'cpu':  # only warmup GPU models
-                im = torch.zeros(*imgsz, dtype=torch.half if self.fp16 else torch.float, device=self.device)  # input
-                for _ in range(2 if self.jit else 1):  #
-                    self.forward(im)  # warmup
+        warmup_types = self.pt, self.jit, self.onnx, self.engine, self.saved_model, self.pb
+        if any(warmup_types) and self.device.type != 'cpu':
+            im = torch.zeros(*imgsz, dtype=torch.half if self.fp16 else torch.float, device=self.device)  # input
+            for _ in range(2 if self.jit else 1):  #
+                self.forward(im)  # warmup
 
     @staticmethod
     def model_type(p='path/to/model.pt'):