@@ -40,6 +40,38 @@ trt_model = tp.from_torch(model, input)
40
40
trt_out = trt_model(input)
41
41
```
42
42
43
+ - 简单的yolo python接口
44
+ - 要求: nvidia驱动 >= 440.33, 操作系统linux x86_64, python37/python38/python39
45
+ - ` pip install trtpy ` , 这是一个实验包,您可以通过pip进行安装,目前提供了yolo支持,删除了repo中的其他项,在不久的将来会提供更新
46
+ - 通过这个指令安装,您就可以使用python接口,而不需要配置cuda、cudnn、tensorRT等系列
47
+ - 内置的版本:CUDA10.2、CUDNN8.2.2.26、TensorRT8.0.1.6、protobuf3.11.4
48
+ ``` python
49
+ import os
50
+ import cv2
51
+ import numpy as np
52
+ import trtpy as tp
53
+
54
+ engine_file = " yolov5s.fp32.trtmodel"
55
+ if not os.path.exists(engine_file):
56
+ tp.compile_onnx_to_file(1 , tp.onnx_hub(" yolov5s" ), engine_file)
57
+
58
+ yolo = tp.Yolo(engine_file, type = tp.YoloType.V5)
59
+ image = cv2.imread(" car.jpg" )
60
+ bboxes = yolo.commit(image).get()
61
+ print (f " { len (bboxes)} objects " )
62
+
63
+ for box in bboxes:
64
+ left, top, right, bottom = map (int , [box.left, box.top, box.right, box.bottom])
65
+ cv2.rectangle(image, (left, top), (right, bottom), tp.random_color(box.class_label), 5 )
66
+
67
+ saveto = " yolov5.car.jpg"
68
+ print (f " Save to { saveto} " )
69
+
70
+ cv2.imwrite(saveto, image)
71
+ cv2.imshow(" result" , image)
72
+ cv2.waitKey()
73
+ ```
74
+
43
75
44
76
## 简介
45
77
1 . 基于tensorRT8.0,C++/Python高级接口
0 commit comments