Research Article

Learning Feature Fusion in Deep Learning-Based Object Detector

Table 1

Symbols used in YOLOv3 loss function.

SymbolDefinition
Number of grid cells in an image
Number of anchor boxes
Confidence score of bounding box in grid cell i
Conditional probability of class c in grid cell i
, , , Location and size of a bounding box
1, if an object appears in grid cell i; 0, otherwise
1, if an object is present in grid cell and the “responsible” bounding box; 0, otherwise
1, if no object is present in grid cell and the “responsible” bounding box; 0, otherwise
1, if predicted label matches the ground truth label; 0, otherwise
Constant (default: 5.0)
Constant (default: 0.5)