MDC_Correlation
Input
- img0:
- Required: yes
- Type: float16
- Description: input tensor with shape[1, 96, 48, 64]
- Restriction: NCHW
- img1:
- Required: yes
- Type: float16
- Description: input tensor with shape[1, 96, 48, 64]
- Restriction: NCHW
Attribute
- pad:
- Required: no
- Type: uint32
- Description: Has the same meaning as that of the convolution operator. Applies to both feature maps.
- Restriction: default = 20
- kernel_size:
- Required: no
- Type: uint32
- Description: convolution kernel size
- Restriction: default = 1
- max_displacement:
- Required: no
- Type: int
- Description: maximum displacement of img1 relative to img0
- Restriction: default = 20
- stride_1:
- Required: no
- Type: int
- Description: stride in pixels of img0
- Restriction: default = 1
- stride_2:
- Required: no
- Type: int
- Description: stride in pixels of img1 within the neighborhood centered around img0
- Restriction: default = 2
Output
featuremap:
- Required: yes
- Type: float16
- Description: similarity between two input feature maps
- Restriction: NCHW
Parent topic: Supported Caffe Operators