Python_demo ONNXruntime 错误

[TJJ120635]

在使用 onnxruntime 示例进行推理时出现了以下错误：

root@ft:~/demo_onnxruntime_lite# python v5lite.py --modelpath v5Lite-e-sim-320.onnx
(6300, 85)
[[1026, 556, 282, 704], [1031, 565, 262, 703], [1031, 552, 270, 722], [1033, 580, 257, 675], [1033, 569, 264, 684], [1022, 563, 299, 667], [1021, 565, 287, 675], [1025, 563, 295, 670], [1019, 564, 289, 677], [1026, 566, 295, 660], [1018, 569, 291, 667], [1031, 556, 268, 684], [995, 528, 311, 741]]
Traceback (most recent call last):
  File "v5lite.py", line 152, in <module>
    srcimg = net.detect(img)
  File "v5lite.py", line 132, in detect
    srcimg = self.postprocess(srcimg, outs, (newh, neww, top, left))
  File "v5lite.py", line 85, in postprocess
    box_index.append(i[0])
IndexError: invalid index to scalar variable.

可能原因是 NMS 方法的返回数组格式问题（版本opencv 4.7.0_py37_armv7l）
没有正确读取到 postprocess 中的 indices = cv2.dnn.NMSBoxes(boxes, confidences, self.confThreshold, self.nmsThreshold) 导致

原代码为：
./python_demo/onnxruntime/v5lite.py 82行

for i in indices:
    box_index.append(i[0])

经过修改如下，可以修复问题：

if indices.ndim == 1:
    for i in indices:
        box_index.append(i)
elif indices.ndim == 2:
    for i in indices:
        box_index.append(i[0])
else:
    raise ValueError("Unexpected shape of indices")