Part Number: TDA2PXEVM
Tool/software: TI-RTOS
Hi,
I am using a model with rectangular kernel size (3x1). Model import was successful but while running on the board, output is not proper. We verified the model on PC and its working fine. Could it be an issue with rectangular kernel size?
Regards
George
Following is the layer in deploy prototext
layer {
name: "ctx_output1/relu_mbox_locc"
type: "Convolution"
bottom: "ctx_output1"
top: "ctx_output1/relu_mbox_locc"
param {
lr_mult: 1.0
decay_mult: 1.0
}
param {
lr_mult: 2.0
decay_mult: 0.0
}
convolution_param {
num_output: 28
bias_term: true
group: 1
weight_filler {
type: "msra"
}
bias_filler {
type: "constant"
value: 0.0
}
pad_h: 1
pad_w: 0
kernel_h: 3
kernel_w: 1
stride_h: 1
stride_w: 1
dilation: 1
}
}
During import, following is the log appearing for this layer:
25, TIDL_ConvolutionLayer , 1, 1 , 1 , 18 , x , x , x , x , x , x , x , 25 , 1 , 256 , 32 , 64 , 1 , 28 , 30 , 62 ,
Expected output size is 32, 64. But it is coming as 30,62. Using netscope tool, it is showing 32, 64. I have attached image also of the same. Why is this happening?
Regards
George
