MATLAB Answers

'importONNXFunction' cannot import yolov3.onnx?

21 views (last 30 days)
cui
cui on 8 Nov 2020
Edited: cui on 16 Mar 2021 at 5:01
When I tried to import 'yolov3.onnx', the following error occurred. How can I solve it?
Extraction code :kuni
%% 导入yolov3.onnx进行预测(推理)
yolov3model = 'yolov3.onnx';
params = importONNXFunction(yolov3model,'yolov3fcn');
%% 推理
img = imread('person.jpg');
img = imresize(img,[416,416]);
img = rescale(img,0,1);
inputImg = permute(img,[3,1,2]);
input_1_01 = reshape(inputImg,[1,size(inputImg)]); % n*c*h*w,[0,1],RGB顺序
image_shape_01 = [1,2];
[yolonms_layer_1_ExpandDims_1_0, yolonms_layer_1_ExpandDims_3_0, yolonms_layer_1_concat_2_0, state] = yolov3fcn(input_1_01, image_shape_01, params,...
'Training',false,...
'InputDataPermutation','none',...
'OutputDataPermutation','none')
Reference to non-existent field 'y3_01'.
Error in yolov3fcn>yolov3Graph2377 (line 1790)
[Vars.TFNodes_yolo_evaluation_layer_1_Shape_3_0, NumDims.TFNodes_yolo_evaluation_layer_1_Shape_3_0] = onnxShape(Vars.y3_01, NumDims.y3_01);
Error in yolov3fcn (line 98)
[yolonms_layer_1_ExpandDims_1_0, yolonms_layer_1_ExpandDims_3_0, yolonms_layer_1_concat_2_0, NumDims.yolonms_layer_1_ExpandDims_1_0, NumDims.yolonms_layer_1_ExpandDims_3_0, NumDims.yolonms_layer_1_concat_2_0, state] =
yolov3Graph2377(input_1_01, image_shape_01, NumDims.input_1_01, NumDims.image_shape_01, Vars, NumDims, Training, params.State);
Error in test1 (line 12)
[yolonms_layer_1_ExpandDims_1_0, yolonms_layer_1_ExpandDims_3_0, yolonms_layer_1_concat_2_0, state] = yolov3fcn(input_1_01, image_shape_01, params,...
MY analyze:
  2 Comments
cui
cui on 24 Feb 2021
I am not really importing and exporting yolov3.onnx, what I want to express is that all onnx model architecture models similar to yolov3.onnx should be imported and exported, such as yolov4.onnx, yolov5.onnx, and other segmentation and detection , The general onnx model of semantics.
I wrote the matlab package of yolov3-v4 a year ago, but it was not supported at that time! My version depends less on the environment and is more efficient!
'importONNXFunction' still needs more improvement!
Update on March 16, 2021, using the latest version of MATLAB 2021a+Deep Learning Toolbox Converter for ONNX Model Format (version 21.1.0) still does not support the import of the above onnx!

Sign in to comment.

Answers (0)

Tags

Products


Release

R2020b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!