Above folder consists of all required files and codes required to convert pretrained models to onnx files so that they can be converted to blob files to be integrated on Oak-D Pipeline.
Note : It is recommended that you place only one blob file in blobs folder and try the corresponding method . For trying a new method you can generate a new blob file and replace it with previous . No method requires more than one blob file to be placed .
Note : Using Midas-Hybrid can reduce the fps by significant amounts
You can directly download the blob files from the given link and place it in Blob folder :
Alternatively , you can generate your own blob files for MiDaS-Small and MiDaS-Hybrid by following the given steps : -
- Run the following script on your terminal window in order to generate blob files for either Midas-small or Midas-Hybrid.
python3 MDE.py
-
Check whether onnx model has been generated in
Onnxfolder. -
Check whether blob model has been generated in
Blobfolder.
You can directly download the blob file for pre-processing from here and place it in Blob folder.
Alternatively , you can generate your own blob file for pre-processing of stereo Images by following the given steps : -
- Run following commands on terminal prompt
python3 Pre_Processing.py