Yolov5FourInputDecodeBox
Input
- x1:
- Required: yes
- Type: float16
- Description: feature_map for 64x downsampling
- Restriction: NCHW
- x2:
- Required: yes
- Type: float16
- Description: feature_map for 32x downsampling
- Restriction: NCHW
- x3:
- Required: yes
- Type: float16
- Description: feature_map for 16x downsampling
- Restriction: NCHW
- x4:
- Required: yes
- Type: float16
- Description: feature_map for 8x downsampling
- Restriction: NCHW
Attribute
- bias_0:
- Required: yes
- Type: float
- Description: the anchor data corresponding to the first input feature_map
- Restriction: six values
- bias_1:
- Required: yes
- Type: float
- Description: the anchor data corresponding to the second input feature_map
- Restriction: six values
- bias_2:
- Required: yes
- Type: float
- Description: the anchor data corresponding to the third input feature_map
- Restriction: six values
- bias_3:
- Required: yes
- Type: float
- Description: the anchor data corresponding to the fourth input feature_map
- Restriction: six values
- ori_size_h:
- Required: yes
- Type: int
- Description: original image height
- ori_size_w:
- Required: yes
- Type: int
- Description: original image width
- anchor:
- Required: yes
- Type: int
- Description: the number of anchors
- Restriction: Must be 3.
- class_num:
- Required: yes
- Type: int
- Description: number of classes
- Restriction: Must be in the range [1, 80].
- thresh_front:
- Required: yes
- Type: float16
- Description: thresholds of confidence value and class score
- Restriction: Must be in the range [0.0, 1.0].
Output
y:
- Required: yes
- Type: float16
- Description: output of parsing the feature_map coordinates and confidence class scores
- Restriction: NCHW
Parent topic: Supported Caffe Operators