1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
| """
YOLOv8 Xilinx Versal导出代码
=============================
"""
import subprocess
import os
def export_yolo_to_versal(
model_name: str = "yolov8n",
weights_path: str = "yolov8n.pt",
output_dir: str = "./versal_export",
quantize: bool = True,
calibration_images: str = "./calibration_images",
):
"""
将YOLOv8模型导出为Versal AI Edge兼容格式
步骤:
1. 导出为ONNX
2. AMD Quark量化 (INT8)
3. Vitis AI编译
Args:
model_name: 模型名称
weights_path: PyTorch权重路径
output_dir: 输出目录
quantize: 是否量化
calibration_images: 量化校准图像目录
"""
os.makedirs(output_dir, exist_ok=True)
print(f"[1/3] 导出ONNX: {weights_path} → {output_dir}/model.onnx")
cmd_onnx = [
"python3", "-c",
f"from ultralytics import YOLO; "
f"model = YOLO('{weights_path}'); "
f"model.export(format='onnx', opset=13, simplify=True)"
]
subprocess.run(cmd_onnx, check=True)
onnx_path = weights_path.replace(".pt", ".onnx")
if not quantize:
print(f"导出完成(无量化): {onnx_path}")
return onnx_path
print(f"[2/3] AMD Quark INT8量化")
cmd_quant = [
"quark", "quantize",
"--input", onnx_path,
"--output", f"{output_dir}/model_int8.onnx",
"--calibration-data", calibration_images,
"--quant-format", "int8",
"--per-channel",
]
subprocess.run(cmd_quant, check=True)
print(f"[3/3] Vitis AI编译")
cmd_compile = [
"vai_q_onnx",
"--model", f"{output_dir}/model_int8.onnx",
"--output_dir", output_dir,
"--arch", "versal_ai_edge_gen2",
]
subprocess.run(cmd_compile, check=True)
compiled_model = f"{output_dir}/model_int8.xmodel"
print(f"✅ 导出完成: {compiled_model}")
print(f"\n=== 性能预估 (Versal AI Edge Gen2) ===")
print(f" 模型: {model_name}")
print(f" 量化: INT8")
print(f" 预计FPS: 200-300 (取决于模型大小)")
print(f" 预计延迟: 3-5ms")
print(f" 预计功耗: 2-5W")
return compiled_model
def export_ims_models():
"""批量导出IMS所需YOLO模型"""
ims_models = [
("yolov8n-face", "yolov8n-face.pt", "人脸检测"),
("yolov8s-pose", "yolov8s-pose.pt", "人体姿态"),
("yolov8n-phone", "yolov8n-phone.pt", "手机检测"),
("yolov8s-seatbelt", "yolov8s-seatbelt.pt", "安全带检测"),
]
for name, weights, desc in ims_models:
print(f"\n{'='*50}")
print(f"导出: {name} ({desc})")
export_yolo_to_versal(
model_name=name,
weights_path=weights,
output_dir=f"./versal_export/{name}",
)
if __name__ == "__main__":
export_ims_models()
|
