does the compute stick with MVNC SDKv2 support the input channel 1(one)?

[lius]

when i run inference my model with input shape of (200, 200, 1), get the wrong result, the out put are all nan.

[JesusE_Intel]

Hi,

Which model are you currently using? The only model that I've seen to work in grayscale is mnist.

Try this model and see if it works for you: https://github.com/BVLC/caffe/blob/master/examples/mnist/lenet.prototxt

You will need to change the input batch parameter from 61 to 1.

Regards,

Jesus

[lius]

Hi，Jesus thank you for your reply. My model is mobile net ssd, we have made modify on the net structure, not the layer,such as self-define layer. I'll try the minist.

[lius]

I have run the lenet(input 1 1 28 28)on movidius and get the right result. But i do really get wrong result of my net, also check the input many times,but get the same result.

[lius]

Hi, Jesus

​ another question, does the mvnc sdk support a convolution kernel which is not square(such 3x1)? I've found that there is a convolution layer which include a kernel like this.

[JesusE_Intel]

The Intel Neural Compute SDK should support non-squared. Could you share your model for me to test? You can post them here or send them over via private message, just click my user name and hit message.

Also, have you tried to use your model using the OpenVINO toolkit?

Regards,

Jesus

[lius]

I'll send you my net in about an hour, and not yet test another sdk.

[lius]

Hi, Jesus the attach include my model and test script. 在 2019-07-23 05:24:54，"Intel Forums" <supportreplies@intel.com> 写道： The Intel Neural Compute SDK should support non-squared. Could you share your model for me to test? You can post them here or send them over via private message, just click my user name and hit message. Also, have you tried to use your model using the OpenVINO toolkit? Regards, Jesus Intel_Jesus (Intel) The Intel Neural Compute SDK should support non-squared. Could you share your model for me to test? You can post them here or send them over via private message, just click my user name and hit message. Also, have you tried to use your model using the OpenVINO toolkit? Regards, Jesus or reply to this email Replying to 3000022673701.563501463246258E12 (Customer) asked a question. Thursday, July 18, 2019 7:14 PM does the compute stick with MVNC SDKv2 support the input channel 1(one)? Intel_Jesus (Intel) The Intel Neural Compute SDK should support non-squared. Could you share your model for me to test? You can post them here or send them over via private message, just click my user name and hit message. Also, have you tried to use your model using the OpenVINO toolkit? Regards, Jesus Monday, July 22, 2019 2:24 PM You're receiving emails when someone "Comments on my posts." To change or turn off Forums email, log in as 300002267370@intel.com.isvc. Are notifications about this post getting annoying? Reply to this email with the word " mute ". For Privacy and Legal notices, please visit these links: http://www.intel.com/privacy, https://www.intel.com/content/www/us/en/legal/terms-of-use.html.

[JesusE_Intel]

I'm not seeing any attachments on your replies, could you please try again?

Regards,

Jesus

[lius]

another question, does the SDK support a model(mobileNet) with non-square input, the last question is non-square weight shape.​

[lius]

I've send you my models and test scripts two days ago, did you get the message?

[JesusE_Intel]

I just saw your messages but no attachments. I've sent you another message to your inbox.

Regards,

Jesus

[JesusE_Intel]

I'm have not tested mobilenet with a non-square input. Is there a particular reason why you are trying to use non-square inputs? I will need to do some test and get back to you.

Regards,

Jesus