1.model_zoo.md

Model Zoo

CASIA-B

Model	[email protected]	[email protected]	[email protected]	mAP.NM	mAP.BG	mAP.CL	Configuration	Input Size
GaitSet(AAAI2019)	95.8(95.0)	90.0(87.2)	75.4(70.4)	85.9	81.4	77.0	gaitset.yaml	64x44
GaitPart(CVPR2020)	96.1(96.2)	90.7(91.5)	78.7(78.7)	84.8	81.1	75.5	gaitpart.yaml	64x44
GLN*(ECCV2020)	96.4(95.6)	93.1(92.0)	81.0(77.2)	89.3	87.0	82.6	gln_phase1.yaml, gln_phase2.yaml	128x88
GaitGL(ICCV2021)	97.4(97.4)	94.5(94.5)	83.8(83.6)	86.2	84.4	79.8	gaitgl.yaml	64x44
GaitBase(CVPR2023)	97.6	94.0	77.4	-	-	-	gaitbase_da_casiab.yaml	64x44

CASIA-B*

Model	[email protected]	[email protected]	[email protected]	Input modality	Configuration
GaitSet(AAAI2019)	92.3	86.1	73.36	Silhouette	gaitset.yaml
GaitPart(CVPR2020)	93.1	86.0	75.1	Silhouette	gaitpart.yaml
GaitGL(ICCV2021)	94.2	90.0	81.4	Silhouette	gaitgl.yaml
GaitEdge(ECCV2022)	97.9	96.1	86.4	RGB, Silhouette	phase2_gaitedge.yaml
GaitBase(CVPR2023)	96.5	91.5	78.0	Silhouette	-

SUSTech1K

Model	[email protected]	[email protected]	[email protected]	[email protected]	[email protected]	[email protected]	[email protected]	[email protected]	[email protected]	[email protected]
PointNet	43.6	37.3	25.7	28.8	19.9	30.1	44.3	27.4	31.3	59.8
PointNet++	55.9	52.2	41.6	49.6	47.8	45.9	54.2	52.5	50.8	82.4
PointTransformer	53.2	48.1	32.1	43.2	39.1	41.8	47.9	47.1	44.4	76.7
SimpleView	72.3	68.8	57.2	63.3	49.2	62.5	79.7	66.5	64.8	85.8
LidarGait(CVPR2023)	91.8	88.6	74.6	89	67.5	80.9	94.5	90.4	86.8	96.1

OUMVLP

Model	Rank@1	mAP	Configuration	Input Size	Inference Time	Model Size
GaitSet(AAAI2019)	87.2(87.1)	92.2	gaitset.yaml	64x44	1m26s	6.31M
GaitPart(CVPR2020)	88.6(88.7)	93.3	gaitpart.yaml	64x44	8m04s	3.78M
GaitGL(ICCV2021)	89.9(89.7)	93.3	gaitgl.yaml	64x44	5m23s	95.62M
GaitBase(CVPR2023)	90.8	-	gaitbase_oumvlp.yaml	64x44	-	-
DeepGaitV2-3D	92.0	-	-	64x44	-	-

GREW

Model	Rank@1	Configuration	Input Size	Inference Time	Model Size
GaitSet(AAAI2019)	48.4	gaitset.yaml	64x44	-	-
GaitPart(CVPR2020)	47.6	gaitpart.yaml	64x44	-	-
GaitGL(ICCV2021)	41.5	gaitgl.yaml	64x44	-	-
GaitGL(BNNeck)(ICCV2021)	51.7	gaitgl.yaml	64x44	-	-
GaitBase(CVPR2023)	60.1	gaitbase_da_grew.yaml	64x44	-	-
DeepGaitV2-3D	79.4	-	64x44	-	-
SwinGait(Transformer-based)	79.3	-	64x44	-	-

Gait3D

Model	Rank@1	Configuration	Input Size	Input modality
GaitSet(AAAI2019)	36.7	-	64x44	Silhouette
GaitPart(CVPR2020)	28.2	-	64x44	Silhouette
GaitGL(ICCV2021)	29.7	-	64x44	Silhouette
SMPLGait(CVPR 2022)	48.3(46.3)	smplgait.yaml	64x44	Silhouette + SMPL
GaitBase(CVPR2023)	64.6	gaitbase_da_gait3d.yaml	64x44	Silhouette
DeepGaitV2-P3D	74.4	-	64x44	-
SwinGait(Transformer-based)	75.0	-	64x44	-

CASIA-E

Model	[email protected]	[email protected]	[email protected]	Input size	Configuration
GaitSet(AAAI2019)	82.54	75.26	62.53	64x44	gaitset_casiae.yaml
GaitPart(CVPR2020)	82.92	74.36	60.48	64x44	gaitpart_casiae.yaml
GaitBase(CVPR2023)	91.59	86.65	74.73	64x44	gaitbase_casiae.yaml

---

The results in the parentheses are mentioned in the papers.

Note:

• The checkpoint files can be found here.
• GaitBase and DeepGaitV2 series are the officially recommended CNN-based baseline models.
• SwinGait series are the officially recommended Transformer-based baseline models.
• All results exclude identical-view cases in CASIA-B(*) and OUMVLP.
• The shown result of GLN is implemented without the compact block.
• The mAP is calculated by the new evaluation protocol proposed in A Comprehensive Study on the Evaluation of Silhouette-based Gait Recognition.