Ask Your Question

Revision history [back]

Does dnn module support bidirectional LSTM?

I want to use the dnn module with a bidirectional RNN layer in Keras. Is this possible? Here's a partial list of supported layers in dnn module, which doesn't include bidirectional. However it states to be partial (Incomplete).

When converting such layer to OpenCV, at inference I get the error message:

...
  feature = network.forward()
cv2.error: OpenCV (4.1.0) {my_dir}\opencv-python\opencv\modules\dnn\src\dnn.cpp:524: error: (-2:Unspecified error) Can't create layer "bidirectional_1/while/Enter_3" of type "Enter" in funtion 'cv::dnn:dnn4_v20190122::LayerData::getLayerInstance'