notesum.ai

Published at October 22

EVC-MF: End-to-end Video Captioning Network with Multi-scale Features

cs.CL

cs.AI

Released Date: October 22, 2024

Authors: Tian-Zi Niu¹, Zhen-Duo Chen¹, Xin Luo¹, Xin-Shun Xu¹

Aff.: ¹Not specified

Arxiv: https://arxiv.org/abs/2410.16624v1

Refer to caption

Method	Features	MSVD				MSR-VTT
Method	Features	B-4	M	R	C	B-4	M	R	C
OSTG (2020) [44]	R200+MR	57.5	36.8	-	92.1	41.9	28.6	-	48.2
OpenBook (2021) [45]	IRV2+C+T	-	-	-	-	42.8	29.3	61.7	52.9
TTA (2021) [46]	R152+C+MR	51.8	35.5	72.4	87.7	41.4	27.7	61.1	46.7
MGRMP (2021) [47]	IRV2+RN	55.8	36.9	74.5	98.5	41.7	28.9	62.1	51.4
TVRD (2022) [48]	IRV2+C+FR	50.5	34.5	71.7	84.3	43.0	28.7	62.2	51.8
vc-HRNAT (2022) [49]	IRV2+I	55.7	36.8	74.1	98.1	42.1	28.0	61.6	48.2
HMN (2022) [50]	IRV2+C+FR	59.2	37.7	75.1	104.0	43.5	29.0	62.7	51.5
HTG+HMG (2023) [51]	R+C	52.7	35.2	72.8	91.4	42.1	28.4	61.6	48.9
VTAR (2023) [52]	IRV2+RN+T	-	-	-	-	44.4	30.0	63.3	56.2
XlanV (2020) [53]	R152+I	-	-	-	-	41.2	28.6	61.5	54.2
SMAN (2022) [54]	IRV2+C+FR	50.2	35.0	71.3	87.7	41.3	28.7	62.1	53.8
SHAN (2022) [55]	IRV2+I	50.9	35.1	72.4	94.5	40.3	28.8	61.2	54.1
CMG (2022) [3]	IRV2+C+a+c	59.5	38.8	76.2	107.3	43.7	29.4	62.8	55.9
SBAT (2020) [56]	IRV2+I	53.1	35.3	72.3	89.5	42.9	28.9	61.5	51.6
STG-KD (2020)[7]	R101+I+FR	52.2	36.9	73.9	93.0	40.5	28.3	60.9	47.1
O2NA (2021) [57]	R101+RN	55.4	37.4	74.5	96.4	41.6	28.5	62.4	51.1
NACF (2021) [38]	R101+RN+c	55.6	36.2	-	96.3	42.0	28.7	-	51.4
LSRT (2022)[58]	IRV2+I+FR	55.6	37.1	73.5	98.5	42.6	28.3	61.0	49.5
SWINBERT (2022)[36]	VS	56.7	40.1	76.2	112.6	42.7	30.4	61.7	54.1
EVC-MF	VS	62.8	41.5	79.0	123.4	45.1	30.4	63.6	57.1