thebears/deepstream_yolo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add dynamic batch-size (ONNX) + Fixes

Browse Source
This commit is contained in:
Marcos Luciano
2023-05-28 13:46:46 -03:00
parent 134960d389
commit 141c0f2fee
20 changed files with 272 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
README.md
View File

@@ -2,16 +2,14 @@
NVIDIA DeepStream SDK 6.2 / 6.1.1 / 6.1 / 6.0.1 / 6.0 configuration for YOLO models
-------------------------------------
### **Big update on DeepStream-Yolo**
-------------------------------------
--------------------------------------------------------------------------------------------------
### Important: please generate the ONNX model and the TensorRT engine again with the updated files
-------------------------------------
--------------------------------------------------------------------------------------------------
### Future updates
* DeepStream tutorials
* Dynamic batch-size
* Updated INT8 calibration
* Support for segmentation models
* Support for classification models
@@ -24,6 +22,7 @@ NVIDIA DeepStream SDK 6.2 / 6.1.1 / 6.1 / 6.0.1 / 6.0 configuration for YOLO mod
* **Support for Darknet YOLO models (YOLOv4, etc) using cfg and weights conversion with GPU post-processing**
* **Support for YOLO-NAS, PPYOLOE+, PPYOLOE, DAMO-YOLO, YOLOX, YOLOR, YOLOv8, YOLOv7, YOLOv6 and YOLOv5 using ONNX conversion with GPU post-processing**
* **Add GPU bbox parser (it is slightly slower than CPU bbox parser on V100 GPU tests)**
* **Dynamic batch-size for ONNX exported models (YOLO-NAS, PPYOLOE+, PPYOLOE, DAMO-YOLO, YOLOX, YOLOR, YOLOv8, YOLOv7, YOLOv6 and YOLOv5)**
##

4
docs/DAMOYOLO.md
View File

@@ -40,7 +40,7 @@ wget https://idstcv.oss-cn-zhangjiakou.aliyuncs.com/DAMO-YOLO/release_model/clea
Generate the ONNX model file (example for DAMO-YOLO-S*)
```
python3 export_damoyolo.py -w damoyolo_tinynasL25_S_477.pth -c configs/damoyolo_tinynasL25_S.py --simplify
python3 export_damoyolo.py -w damoyolo_tinynasL25_S_477.pth -c configs/damoyolo_tinynasL25_S.py --simplify --dynamic
```
**NOTE**: To change the inference size (defaut: 640)
@@ -66,7 +66,7 @@ or
#### 5. Copy generated files
Copy the generated ONNX model file to the `DeepStream-Yolo` folder.
Copy the generated ONNX model file and labels.txt file (if generated) to the `DeepStream-Yolo` folder.
##

2
docs/PPYOLOE.md
View File

@@ -43,7 +43,7 @@ python3 export_ppyoloe.py -w ppyoloe_plus_crn_s_80e_coco.pdparams -c configs/ppy
#### 5. Copy generated files
Copy the generated ONNX model file to the `DeepStream-Yolo` folder.
Copy the generated ONNX model file and labels.txt file (if generated) to the `DeepStream-Yolo` folder.
##

4
docs/YOLONAS.md
View File

@@ -43,7 +43,7 @@ wget https://sghub.deci.ai/models/yolo_nas_s_coco.pth
Generate the ONNX model file (example for YOLO-NAS S)
```
python3 export_yolonas.py -m yolo_nas_s -w yolo_nas_s_coco.pth --simplify
python3 export_yolonas.py -m yolo_nas_s -w yolo_nas_s_coco.pth --simplify --dynamic
```
**NOTE**: Model names
@@ -85,7 +85,7 @@ or
-s 1280 1280
```
#### 5. Copy generated files
#### 5. Copy generated file
Copy the generated ONNX model file to the `DeepStream-Yolo` folder.

6
docs/YOLOR.md
View File

@@ -44,7 +44,7 @@ Generate the ONNX model file
Example for YOLOR-CSP
```
python3 export_yolor.py -w yolor_csp.pt -c cfg/yolor_csp.cfg --simplify
python3 export_yolor.py -w yolor_csp.pt -c cfg/yolor_csp.cfg --simplify --dynamic
```
- Paper branch
@@ -52,7 +52,7 @@ Generate the ONNX model file
Example for YOLOR-P6
```
python3 export_yolor.py -w yolor-p6.pt --simplify
python3 export_yolor.py -w yolor-p6.pt --simplify --dynamic
```
**NOTE**: To convert a P6 model
@@ -84,7 +84,7 @@ or
#### 5. Copy generated files
Copy the generated ONNX model file to the `DeepStream-Yolo` folder
Copy the generated ONNX model file and labels.txt file (if generated) to the `DeepStream-Yolo` folder
##

4
docs/YOLOX.md
View File

@@ -43,10 +43,10 @@ wget https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yo
Generate the ONNX model file (example for YOLOX-s)
```
python3 export_yolox.py -w yolox_s.pth -c exps/default/yolox_s.py --simplify
python3 export_yolox.py -w yolox_s.pth -c exps/default/yolox_s.py --simplify --dynamic
```
#### 5. Copy generated files
#### 5. Copy generated file
Copy the generated ONNX model file to the `DeepStream-Yolo` folder.

4
docs/YOLOv5.md
View File

@@ -44,7 +44,7 @@ wget https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5s.pt
Generate the ONNX model file (example for YOLOv5s)
```
python3 export_yoloV5.py -w yolov5s.pt --simplify
python3 export_yoloV5.py -w yolov5s.pt --simplify --dynamic
```
**NOTE**: To convert a P6 model
@@ -76,7 +76,7 @@ or
#### 5. Copy generated files
Copy the generated ONNX model file to the `DeepStream-Yolo` folder.
Copy the generated ONNX model file and labels.txt file (if generated) to the `DeepStream-Yolo` folder.
##

4
docs/YOLOv6.md
View File

@@ -44,7 +44,7 @@ wget https://github.com/meituan/YOLOv6/releases/download/0.4.0/yolov6s.pt
Generate the ONNX model file (example for YOLOv6-S 4.0)
```
python3 export_yoloV6.py -w yolov6s.pt --simplify
python3 export_yoloV6.py -w yolov6s.pt --simplify --dynamic
```
**NOTE**: To convert a P6 model
@@ -74,7 +74,7 @@ or
-s 1280 1280
```
#### 5. Copy generated files
#### 5. Copy generated file
Copy the generated ONNX model file to the `DeepStream-Yolo` folder.

4
docs/YOLOv7.md
View File

@@ -46,7 +46,7 @@ wget https://github.com/WongKinYiu/yolov7/releases/download/v0.1/yolov7.pt
Generate the ONNX model file (example for YOLOv7)
```
python3 export_yoloV7.py -w yolov7.pt --simplify
python3 export_yoloV7.py -w yolov7.pt --simplify --dynamic
```
**NOTE**: To convert a P6 model
@@ -78,7 +78,7 @@ or
#### 6. Copy generated files
Copy the generated ONNX model file to the `DeepStream-Yolo` folder.
Copy the generated ONNX model file and labels.txt file (if generated) to the `DeepStream-Yolo` folder.
##

4
docs/YOLOv8.md
View File

@@ -43,7 +43,7 @@ wget https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8s.pt
Generate the ONNX model file (example for YOLOv8s)
```
python3 export_yoloV8.py -w yolov8s.pt --simplify
python3 export_yoloV8.py -w yolov8s.pt --simplify --dynamic
```
**NOTE**: To change the inference size (defaut: 640)
@@ -69,7 +69,7 @@ or
#### 5. Copy generated files
Copy the generated ONNX model file to the `DeepStream-Yolo` folder.
Copy the generated ONNX model file and labels.txt file (if generated) to the `DeepStream-Yolo` folder.
##

29
utils/export_damoyolo.py
View File

@@ -46,9 +46,21 @@ def damoyolo_export(weights, config_file, device):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening DAMO-YOLO model')
device = torch.device('cpu')
cfg, model = damoyolo_export(args.weights, args.config, device)
if len(cfg.dataset['class_names']) > 0:
print('Creating labels.txt file')
f = open('labels.txt', 'w')
for name in cfg.dataset['class_names']:
f.write(name + '\n')
f.close()
model = nn.Sequential(model, DeepStreamOutput())
img_size = args.size * 2 if len(args.size) == 1 else args.size
@@ -56,15 +68,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = cfg.miscs['exp_name'] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('Exporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream DAMO-YOLO conversion')
@@ -73,6 +99,7 @@ def parse_args():
parser.add_argument('-s', '--size', nargs='+', type=int, default=[640], help='Inference size [H,W] (default [640])')
parser.add_argument('--opset', type=int, default=11, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if not os.path.isfile(args.weights):
raise SystemExit('Invalid weights file')

20
utils/export_ppyoloe.py
View File

@@ -8,6 +8,7 @@ from ppdet.utils.check import check_version, check_config
from ppdet.utils.cli import ArgsParser
from ppdet.engine import Trainer
from ppdet.slim import build_slim_model
from ppdet.data.source.category import get_categories
class DeepStreamOutput(nn.Layer):
@@ -39,13 +40,26 @@ def ppyoloe_export(FLAGS):
os.makedirs('.tmp')
static_model, _ = trainer._get_infer_cfg_and_input_spec('.tmp')
os.system('rm -r .tmp')
return cfg, static_model
return trainer.cfg, static_model
def main(FLAGS):
print('\nStarting: %s' % FLAGS.weights)
print('\nOpening PPYOLOE model\n')
paddle.set_device('cpu')
cfg, model = ppyoloe_export(FLAGS)
anno_file = cfg['TestDataset'].get_anno()
if os.path.isfile(anno_file):
_, catid2name = get_categories(cfg['metric'], anno_file, 'detection_arch')
print('\nCreating labels.txt file')
f = open('labels.txt', 'w')
for name in catid2name.values():
f.write(str(name) + '\n')
f.close()
model = nn.Sequential(model, DeepStreamOutput())
img_size = [cfg.eval_height, cfg.eval_width]
@@ -55,14 +69,18 @@ def main(FLAGS):
onnx_input_im['scale_factor'] = paddle.static.InputSpec(shape=[None, 2], dtype='float32', name='scale_factor')
onnx_output_file = cfg.filename + '.onnx'
print('\nExporting the model to ONNX\n')
paddle.onnx.export(model, cfg.filename, input_spec=[onnx_input_im], opset_version=FLAGS.opset)
if FLAGS.simplify:
print('\nSimplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('\nDone: %s\n' % onnx_output_file)
def parse_args():
parser = ArgsParser()

29
utils/export_yoloV5.py
View File

@@ -41,9 +41,21 @@ def yolov5_export(weights, device):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening YOLOv5 model\n')
device = select_device('cpu')
model = yolov5_export(args.weights, device)
if len(model.names.keys()) > 0:
print('\nCreating labels.txt file')
f = open('labels.txt', 'w')
for name in model.names.values():
f.write(name + '\n')
f.close()
model = nn.Sequential(model, DeepStreamOutput())
img_size = args.size * 2 if len(args.size) == 1 else args.size
@@ -54,15 +66,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = os.path.basename(args.weights).split('.pt')[0] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('\nExporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream YOLOv5 conversion')
@@ -71,6 +97,7 @@ def parse_args():
parser.add_argument('--p6', action='store_true', help='P6 model')
parser.add_argument('--opset', type=int, default=17, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if not os.path.isfile(args.weights):
raise SystemExit('Invalid weights file')

22
utils/export_yoloV6.py
View File

@@ -51,6 +51,11 @@ def yolov6_export(weights, device):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening YOLOv6 model\n')
device = torch.device('cpu')
model = yolov6_export(args.weights, device)
@@ -64,15 +69,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = os.path.basename(args.weights).split('.pt')[0] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('\nExporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream YOLOv6 conversion')
@@ -81,6 +100,7 @@ def parse_args():
parser.add_argument('--p6', action='store_true', help='P6 model')
parser.add_argument('--opset', type=int, default=13, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if not os.path.isfile(args.weights):
raise SystemExit('Invalid weights file')

29
utils/export_yoloV7.py
View File

@@ -45,9 +45,21 @@ def yolov7_export(weights, device):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening YOLOv7 model\n')
device = select_device('cpu')
model = yolov7_export(args.weights, device)
if len(model.names) > 0:
print('\nCreating labels.txt file')
f = open('labels.txt', 'w')
for name in model.names:
f.write(name + '\n')
f.close()
model = nn.Sequential(model, DeepStreamOutput())
img_size = args.size * 2 if len(args.size) == 1 else args.size
@@ -58,15 +70,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = os.path.basename(args.weights).split('.pt')[0] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('\nExporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream YOLOv7 conversion')
@@ -75,6 +101,7 @@ def parse_args():
parser.add_argument('--p6', action='store_true', help='P6 model')
parser.add_argument('--opset', type=int, default=12, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if not os.path.isfile(args.weights):
raise SystemExit('Invalid weights file')

29
utils/export_yoloV7_u6.py
View File

@@ -40,9 +40,21 @@ def yolov7_u6_export(weights, device):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening YOLOv7_u6 model\n')
device = select_device('cpu')
model = yolov7_u6_export(args.weights, device)
if len(model.names.keys()) > 0:
print('\nCreating labels.txt file')
f = open('labels.txt', 'w')
for name in model.names.values():
f.write(name + '\n')
f.close()
model = nn.Sequential(model, DeepStreamOutput())
img_size = args.size * 2 if len(args.size) == 1 else args.size
@@ -50,15 +62,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = os.path.basename(args.weights).split('.pt')[0] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('\nExporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream YOLOv7-u6 conversion')
@@ -66,6 +92,7 @@ def parse_args():
parser.add_argument('-s', '--size', nargs='+', type=int, default=[640], help='Inference size [H,W] (default [640])')
parser.add_argument('--opset', type=int, default=12, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if not os.path.isfile(args.weights):
raise SystemExit('Invalid weights file')

29
utils/export_yoloV8.py
View File

@@ -48,9 +48,21 @@ def yolov8_export(weights, device):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening YOLOv8 model\n')
device = select_device('cpu')
model = yolov8_export(args.weights, device)
if len(model.names.keys()) > 0:
print('\nCreating labels.txt file')
f = open('labels.txt', 'w')
for name in model.names.values():
f.write(name + '\n')
f.close()
model = nn.Sequential(model, DeepStreamOutput())
img_size = args.size * 2 if len(args.size) == 1 else args.size
@@ -58,15 +70,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = os.path.basename(args.weights).split('.pt')[0] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('\nExporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream YOLOv8 conversion')
@@ -74,6 +100,7 @@ def parse_args():
parser.add_argument('-s', '--size', nargs='+', type=int, default=[640], help='Inference size [H,W] (default [640])')
parser.add_argument('--opset', type=int, default=16, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if not os.path.isfile(args.weights):
raise SystemExit('Invalid weights file')

22
utils/export_yolonas.py
View File

@@ -34,6 +34,11 @@ def yolonas_export(model_name, weights, num_classes, size):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening YOLO-NAS model\n')
device = torch.device('cpu')
model = yolonas_export(args.model, args.weights, args.classes, args.size)
@@ -44,15 +49,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = os.path.basename(args.weights).split('.pt')[0] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('\nExporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream YOLO-NAS conversion')
@@ -62,6 +81,7 @@ def parse_args():
parser.add_argument('-s', '--size', nargs='+', type=int, default=[640], help='Inference size [H,W] (default [640])')
parser.add_argument('--opset', type=int, default=14, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if args.model == '':
raise SystemExit('Invalid model name')

29
utils/export_yolor.py
View File

@@ -57,9 +57,21 @@ def yolor_export(weights, cfg, size, device):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening YOLOR model\n')
device = torch.device('cpu')
model = yolor_export(args.weights, args.cfg, args.size, device)
if hasattr(model, 'names') and len(model.names) > 0:
print('\nCreating labels.txt file')
f = open('labels.txt', 'w')
for name in model.names:
f.write(name + '\n')
f.close()
model = nn.Sequential(model, DeepStreamOutput())
img_size = args.size * 2 if len(args.size) == 1 else args.size
@@ -70,15 +82,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = os.path.basename(args.weights).split('.pt')[0] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('\nExporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream YOLOR conversion')
@@ -88,6 +114,7 @@ def parse_args():
parser.add_argument('--p6', action='store_true', help='P6 model')
parser.add_argument('--opset', type=int, default=12, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if not os.path.isfile(args.weights):
raise SystemExit('Invalid weights file')

22
utils/export_yolox.py
View File

@@ -42,6 +42,11 @@ def yolox_export(weights, exp_file):
def main(args):
suppress_warnings()
print('\nStarting: %s' % args.weights)
print('Opening YOLOX model')
device = torch.device('cpu')
model, exp = yolox_export(args.weights, args.exp)
@@ -52,15 +57,29 @@ def main(args):
onnx_input_im = torch.zeros(1, 3, *img_size).to(device)
onnx_output_file = os.path.basename(args.weights).split('.pt')[0] + '.onnx'
dynamic_axes = {
'input': {
0: 'batch'
},
'output': {
0: 'batch'
}
}
print('Exporting the model to ONNX')
torch.onnx.export(model, onnx_input_im, onnx_output_file, verbose=False, opset_version=args.opset,
do_constant_folding=True, input_names=['input'], output_names=['output'], dynamic_axes=None)
do_constant_folding=True, input_names=['input'], output_names=['output'],
dynamic_axes=dynamic_axes if args.dynamic else None)
if args.simplify:
print('Simplifying the ONNX model')
import onnxsim
model_onnx = onnx.load(onnx_output_file)
model_onnx, _ = onnxsim.simplify(model_onnx)
onnx.save(model_onnx, onnx_output_file)
print('Done: %s\n' % onnx_output_file)
def parse_args():
parser = argparse.ArgumentParser(description='DeepStream YOLOX conversion')
@@ -68,6 +87,7 @@ def parse_args():
parser.add_argument('-c', '--exp', required=True, help='Input exp (.py) file path (required)')
parser.add_argument('--opset', type=int, default=11, help='ONNX opset version')
parser.add_argument('--simplify', action='store_true', help='ONNX simplify model')
parser.add_argument('--dynamic', action='store_true', help='Dynamic batch-size')
args = parser.parse_args()
if not os.path.isfile(args.weights):
raise SystemExit('Invalid weights file')