search

BBuf / cv_tools

62 stars 17 forks source link

I use yolov3.trt, not Yolo tiny reasoning, why there is no result? What are the changes? I made the following changes #1

Open kongxm889 opened 4 years ago

kongxm889 commented 4 years ago

string onnxFile = "F:/TensorRT-6.0.1.5/data/v3tiny/yolov3-tiny.onnx"; string engineFile = "F:/TensorRT-6.0.1.5/data/v3tiny/yolov3-tiny.trt";

vector labels = { "bad" };

vector<vector > output_shape = { { 1, 18, 13, 13 },{ 1, 18, 26, 26 } ,{1,18,52,52}}; vector<vector > g_masks = {{6,7,8}, { 3, 4, 5 },{ 0, 1, 2 } }; vector<vector > g_anchors = { { 10, 14 },{18,16},{ 23, 27 },{ 37, 58 },{40,56},{66, 68},{ 81, 82 },{ 135, 169 },{ 344, 319} };

BBuf commented 4 years ago

Please see this：https://github.com/BBuf/yolov3-tiny-onnx-TensorRT

kongxm889 commented 4 years ago

请问你可以改个yolov3.trt的推理C++代码吗，我在tiny上没有改对。我用的是大yolo，3个输出的那个网络

BBuf commented 4 years ago

Please see this：https://github.com/BBuf/yolov3-tiny-onnx-TensorRT

kongxm889 commented 4 years ago

mance or even cause errors. prepare image take: 129.974 ms. Inference take: 7.2566 ms. Postprocess take: 3153.07 ms. 0.92241, 221.538, 458.154, 0.00018106, 5.26106e-05 0.771616, 541.538, 433.539, 0.000305597, 2.88282e-05 0.751903, 196.931, 248.922, 0.00325906, 0.022265 0.665877, 184.626, 248.922, 0.0105204, 0.0374328 0.586902, 603.114, 310.721, 0.00889372, 0.00443592 0.57666, 123.065, 285.884, 0.160754, 0.107473 0.576477, 110.748, 285.861, 0.112664, 0.042417 0.571044, 98.4042, 285.852, 0.149839, 0.0392293 0.56572, 209.235, 248.931, 0.00352389, 0.0237938 0.564665, 270.77, 310.462, 1.16625e-05, 0.000262268 0.544211, 283.16, 248.916, 0.0718938, 0.0699337 0.54115, 86.0993, 285.862, 0.12917, 0.0359988 0.539281, 295.467, 248.751, 0.574393, 0.528999 0.534978, 516.925, 273.552, 0.0032337, 0.00539271 0.50896, 61.3222, 371.985, 0.450847, 0.0464043

这是什么原因

BBuf commented 4 years ago

啥。。

kongxm889 commented 4 years ago

这是我测试图像输出的结果，宽高输出不对