I have a private model, and input shape is [-1,-1,-1,3], it can be handled by tensorflw tools graph_transform.
when i use "snpe-tensorflow-to-dlc", found "--input_dim" must is a 3dim shape like[x,y,z],and must fixed-size input!!!
What should i do, if i need a 4dim input shape and it is NOT fixed-size? (I think this snpe convert tools is a graph_transform wrapper, and graph_transform can handle 4dim and Non-Fixed_Size input placeholder)
I have a similar question. For the object detection model
https://github.com/tensorflow/models/tree/master/research/object_detection
i have input as "image_tesnor" with shape [1,None,None,3]
How to generate dlc model for this type of tensorflow network ?
it occures to me,have you solve this problem with input_dims is no-fixed_size
any one solve this issue!!!
please help me to find the solution