You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I know NCNN did not support 3D convolution in 2019. (#889)
But I still wanna give it a try.
At first, I have a PyTorch 3D convolution model.
PyTorch model -> ONNX model is fine
ONNX model -> simplified ONNX model is fine
simplified ONNX model -> NCNN model is fine (The model convertor did not show any error message like not support 3D convolution)
However, when I call ncnn::Extractor ex = resnet3d.create_extractor();
an exception "Integer division by zero" popped out.
I wonder that is the exception due to NCNN does not support 3D convolution. If NCNN does not support 3D convolution, will NCNN support 3D convolution in the near feature?
error log | 日志或报错信息 | ログ
I know NCNN did not support 3D convolution in 2019. (#889)
But I still wanna give it a try.
At first, I have a PyTorch 3D convolution model.
PyTorch model -> ONNX model is fine
ONNX model -> simplified ONNX model is fine
simplified ONNX model -> NCNN model is fine (The model convertor did not show any error message like not support 3D convolution)
However, when I call
ncnn::Extractor ex = resnet3d.create_extractor();
an exception "Integer division by zero" popped out.
I wonder that is the exception due to NCNN does not support 3D convolution.
If NCNN does not support 3D convolution, will NCNN support 3D convolution in the near feature?
Thanks
model | 模型 | モデル
how to reproduce | 复现步骤 | 再現方法
ncnn::Net resnet3d;
resnet3d.load_param("resnet3Dmodel.param");
resnet3d.load_model("resnet3Dmodel.bin");
ncnn::Mat in = ncnn::Mat::from_pixels_resize(image.data, ncnn::Mat::PIXEL_BGR, image.cols, image.rows, image.cols, image.rows);
const float mean_vals[3] = { 0, 0, 0 };
const float norm_vals[3] = { 1.0 / 255, 1.0 / 255, 1.0 / 255 };
in.substract_mean_normalize(mean_vals, norm_vals);
ncnn::Mat out;
ncnn::Extractor ex = resnet3d.create_extractor();
ex.input("input", in);
ex.extract("output", out);
The text was updated successfully, but these errors were encountered: