- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Guys in Intel,
I just wanted to find out the reason NCS2 does not produce same outputs as expected, and made a simple experiment.
I created a very simple IR model which contains only one layer, like this( I also attach the model)
<?xml version="1.0" ?> <net batch="1" name="areca" version="4"> <layers> <layer id="0" name="inputs" precision="FP16" type="Input"> <output> <port id = "0"> <dim>1</dim> <dim>3</dim> <dim>416</dim> <dim>416</dim> </port> </output> </layer> <layer id="1" name="layer01.convolution" precision="FP16" type="Convolution"> <data auto_pad="same_upper" dilations="1,1" group="1" kernel="3,3" output="16" pads_begin="1,1" pads_end="1,1" strides="1,1"/> <input> <port id="0"> <dim>1</dim> <dim>3</dim> <dim>416</dim> <dim>416</dim> </port> </input> <output> <port id="3"> <dim>1</dim> <dim>16</dim> <dim>416</dim> <dim>416</dim> </port> </output> <blobs> <weights offset="0" size="864" /> <biases offset="864" size="32" /> </blobs> </layer> </layers> <edges> <edge from-layer="0" from-port="0" to-layer="1" to-port="0"/> </edges> <meta_data> </meta_data> </net>
and load image data from input.bin (in the attachment), the NCS produce totally different results compared to that I produce in nVidia GeForce 1060( with datatype FP16), as below:
This also leads to a totally different detection result on GPU
I also attach the visual studio 2015 project and the source file for your reference.
Please help me to fix this problem, many thanks.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Dear Tsin, Ross,
Can you kindly try the OpenVino Python Sample object_detection_demo_yolov3_async on this image ? That sample accepts images for -i . Please let me know what happens.
Thanks,
Shubha
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page