search

PINTO0309 / MobileNet-SSDLite-RealSense-TF

RaspberryPi3(Raspbian Stretch) + MobileNetv2-SSDLite(Tensorflow/MobileNetv2SSDLite) + RealSense D435 + Tensorflow1.11.0 + without Neural Compute Stick(NCS)
https://qiita.com/PINTO
MIT License
51 stars 16 forks source link

Neural Compute Stick can support TF?I want apply NCS to TF #2

Closed mk123qwe closed 6 years ago

mk123qwe commented 6 years ago

your title is “ without Neural Compute Stick(NCS)” .Neural Compute Stick can not support TF? I am only familiar with TF.I want apply NCS to TF.could you give me some suggestions？

PINTO0309 commented 6 years ago

@mk123qwe https://ncsforum.movidius.com/discussion/746/is-it-possible-to-use-tensorflow-ssd-mobilenet-on-ncs

As time is wasted if you are using an ARM processor, we encourage you to use "OpenVINO" on x86_64 processor hardware. Intel does not seem to have a roadmap corresponding to tensorflow.

mk123qwe commented 6 years ago

Thank you for your answer. I find information that NCS support some caffe model.I prefer to gpu if using x86_64 processor .I pay more attention to ARM with deeplearning.

PINTO0309 commented 6 years ago

@mk123qwe Conversion from tensorflow to caffe is tough because there are a lot of layers unsupported. I abandoned it already. If you succeeded, please tell me. good luck.

PINTO0309 commented 6 years ago

Since one week passed, I will close "issue" once. Please open "issue" again if necessary.