Table 3.

Infrastructure of the backbone and neck parts of the teacher and student models. † indicates that this is a special convolutional layer [55].

Model	1 × 1 Conv	3 × 3 Conv	Depthwise Conv †	Pointwise Conv †
YOLOR [38]	86	53	0	0
YOLOR-Light-v1	23	21	0	0
YOLOR-Light-v2	53	32	0	0
Mobile-YOLOR-v1	30	10	17	34
Mobile-YOLOR-v2	52	10	17	34