Problem with loading ONNX networks with Resize layer #18300
Description
System information (version)
- OpenCV => 4.4.0-dev or 3.4.11-pre
- Operating System / Platform => Ubuntu 18.04
- Compiler => gcc 7.5.0
Detailed description
OpenCV has the following problem with loading networks with Resize layer:
cv2.error: OpenCV(3.4.11-pre) /home/ssl/Documents/projects/OpenCV/opencv/modules/dnn/src/onnx/onnx_importer.cpp:1424: error: (-2:Unspecified error) in function 'void cv::dnn::experimental_dnn_34_v18::ONNXImporter::populateNet(cv::dnn::experimental_dnn_34_v18::Net)'
(expected: 'shapes.depth() == CV_32S'), where
'shapes.depth()' is 5 (CV_32FC1)
must be equal to
'CV_32S' is 4 (CV_32SC1)
It prevents OpenCV from loading various networks from Segmentation Models Pytorch (https://github.com/qubvel/segmentation_models.pytorch) or Human Segmentation Pytorch (https://github.com/thuyngch/Human-Segmentation-PyTorch)
Steps to reproduce
reproducer_resize.py
import cv2
import torch
import torch.nn as nn
import torch.nn.functional as F
class Conv2dReLU(nn.Sequential):
def __init__(
self,
in_channels,
out_channels,
kernel_size,
padding=0,
stride=1,
use_batchnorm=True,
):
conv = nn.Conv2d(
in_channels,
out_channels,
kernel_size,
stride=stride,
padding=padding,
bias=not (use_batchnorm),
)
relu = nn.ReLU(inplace=True)
if use_batchnorm and use_batchnorm != "inplace":
bn = nn.BatchNorm2d(out_channels)
else:
bn = nn.Identity()
super(Conv2dReLU, self).__init__(conv, bn, relu)
class Attention(nn.Module):
def __init__(self, name, **params):
super().__init__()
if name is None:
self.attention = nn.Identity(**params)
else:
raise ValueError("Attention {} is not implemented".format(name))
def forward(self, x):
return self.attention(x)
class DecoderBlock(nn.Module):
def __init__(
self,
in_channels,
skip_channels,
out_channels,
use_batchnorm=True,
attention_type=None,
):
super().__init__()
self.conv1 = Conv2dReLU(
in_channels + skip_channels,
out_channels,
kernel_size=3,
padding=1,
use_batchnorm=use_batchnorm,
)
self.conv2 = Conv2dReLU(
out_channels,
out_channels,
kernel_size=3,
padding=1,
use_batchnorm=use_batchnorm,
)
def forward(self, x, skip=None):
x = F.interpolate(x, scale_factor=2, mode="nearest")
if skip is not None:
x = torch.cat([x, skip], dim=1)
x = self.conv1(x)
x = self.conv2(x)
return x
model = DecoderBlock(2, 0, 2)
x = torch.rand(1, 2, 10, 10)
model.eval()
print(cv2.__version__)
torch.onnx.export(model, x, "resize.onnx", opset_version=11)
opencv_net = cv2.dnn.readNetFromONNX("resize.onnx")
Issue submission checklist
- I report the issue, it's not a question
- I checked the problem with documentation, FAQ, open issues,
answers.opencv.org, Stack Overflow, etc and have not found solution - I updated to latest OpenCV version and the issue is still there
- There is reproducer code and related data files: videos, images, onnx, etc