Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.

How to crop with NCS

idata
Employee
741 Views

Hi everybody, i have changed all rectangular convolution in my model to square ones(padding with zeros) and i have padded my inputs properly.

 

After this i have cropped some of the garbage and finally the networks have the same performance… but i used DummyData to pass the reference blob to the Crop-Layer, which is not supported, so i can i do something similar without using DummyData?

 

name: "UNIPINET" input:"data" input_shape{ dim: 1 dim: 1 dim: 63 dim: 13 } layer { name: "crop1" type: "DummyData" top: "crop1" dummy_data_param{ shape: { dim: 1 dim: 1 dim:49 dim:11 } data_filler { type: "constant" value: 0 } } } layer { name: "crop2" type: "DummyData" top: "crop2" dummy_data_param{ shape: { dim: 1 dim: 16 dim: 40 dim: 9 } data_filler { type: "constant" value: 0 } } } layer { name: "crop3" type: "DummyData" top: "crop3" dummy_data_param{ shape: { dim: 1 dim: 64 dim: 36 dim: 7 } data_filler { type: "constant" value: 0 } } } layer { name: "conv1/dw" type: "Convolution" bottom: "data" top: "conv1/dw" param { lr_mult: 1 decay_mult: 1 } convolution_param { num_output: 1 bias_term: false kernel_h: 15 kernel_w: 15 pad_w: 12 group: 1 #engine: CAFFE stride: 1 weight_filler { type: "constant" value: 0 } } } layer { name: "cropL1" type: "Crop" bottom: "conv1/dw" bottom: "crop1" top: "conv1/dw/cropped" crop_param { axis: -1 offset: 12 } } layer { name: "conv1/sep" type: "Convolution" bottom: "conv1/dw/cropped" top: "conv1/sep" param { lr_mult: 1 decay_mult: 1 } convolution_param { num_output: 16 bias_term: true pad: 0 kernel_size: 1 stride: 1 weight_filler { type: "msra" } } }
0 Kudos
0 Replies
Reply