Visual-Text Reference Pretraining Model for Image Captioning

Table 1

Comparisons with state-of-the-art single-model approaches on MSCOCO karpathy test split.

Approach	B@1	B@2	B@3	B@4	M	R	C	S
ADAPTIVE [24]	74.8	58.4	44.4	33.6	26.4	55.0	104.2	19.7
UP-DOWN [45]	80.2	64.1	49.1	36.3	27.7	56.9	120.1	21.4
CAVP [46]	80.1	64.7	50.0	38.6	28.3	58.9	126.3	21.6
SGAE [26]	80.6	65.0	50.1	39.0	28.4	58.9	129.1	22.2
ORT [47]	80.8	—	—	38.6	28.7	58.4	128.3	22.6
AOANET [48]	81.0	65.8	—	38.9	29.2	58.8	129.8	22.4
U-VLP [7]	—	—	—	39.5	29.3	—	129.3	23.2
NG-SAN [29]	80.8	65.4	50.8	39.9	29.3	59.2	132.1	23.3
ASG [27]	—	—	—	23.0	24.5	50.1	204.2	42.1
VTR-PTM (ours)	82.9	67.3	53.4	40.9	30.9	61.5	130.2	28.5