地瓜官网
地瓜开发者社区
地瓜官网
地瓜开发者社区
中文

S100 模型性能Benchmark

  • 测试开发板：S100

  • 模型来源：OE包内 samples/ucp_tutorial/dnn/ai_benchmark 路径下的模型

  • 运行环境：Linux

模型精度

MODEL NAMEINPUT SIZEACCURACYDataset
ResNet50

1x3x224x224

Top1:

0.7704(FLOAT)/0.7665(INT8)

ImageNet
GoogleNet

1x3x224x224

Top1:

0.7018(FLOAT)/0.6998(INT8)

ImageNet
EfficientNet_Lite1

1x240x240x3

Top1:

0.7652(FLOAT)/0.7614(INT8)

ImageNet
EfficientNet_Lite2

1x260x260x3

Top1:

0.7734(FLOAT)/0.7697(INT8)

ImageNet
EfficientNet_Lite3

1x280x280x3

Top1:

0.7917(FLOAT)/0.7896(INT8)

ImageNet
EfficientNet_Lite4

1x300x300x3

Top1:

0.8063(FLOAT)/0.8043(INT8)

ImageNet
Vargconvnet

1x3x224x224

Top1:

0.7793(FLOAT)/0.7770(INT8)

ImageNet
Efficientnasnet_m

1x3x300x300

Top1:

0.7935(FLOAT)/0.7923(INT8)

ImageNet
Efficientnasnet_s

1x3x280x280

Top1:

0.7441(FLOAT)/0.7524(INT8)

ImageNet
ResNet18

1x3x224x224

Top1:

0.7170(FLOAT)/0.7157(INT8)

ImageNet
YOLOv2_Darknet19

1x3x608x608

[IoU=0.50:0.95]=

0.2760(FLOAT)/0.2700(INT8)

COCO
YOLOv3_Darknet53

1x3x416x416

[IoU=0.50:0.95]=

0.3370(FLOAT)/0.3360(INT8)

COCO
YOLOv5x_v2.0

1x3x672x672

[IoU=0.50:0.95]=

0.4810(FLOAT)/0.4670(INT8)

COCO
Centernet_resnet101

1x3x512x512

[IoU=0.50:0.95]=

0.3420(FLOAT)/0.3270(INT8)

COCO
YOLOv3_VargDarknet

1x3x416x416

[IoU=0.50:0.95]=

0.3280(FLOAT)/0.3270(INT8)

COCO
Deeplabv3plus_efficientnetb0

1x3x1024x2048

mIoU:

0.7630(FLOAT)/0.7571(INT8)

Cityscapes
Fastscnn_efficientnetb0

1x3x1024x2048

mIoU:

0.6997(FLOAT)/0.6914(INT8)

Cityscapes
Deeplabv3plus_efficientnetm1

1x3x1024x2048

mIoU:

0.7794(FLOAT)/0.7754(INT8)

Cityscapes
Deeplabv3plus_efficientnetm2

1x3x1024x2048

mIoU:

0.7882(FLOAT)/0.7854(INT8)

Cityscapes
Bev_gkt_mixvargenet_multitask

image:

6x3x512x960

points(0-8):

6x64x64x2

NDS:

0.2810(FLOAT)/0.2781(INT8)

MeanIOU:

0.4852(FLOAT)/0.4839(INT8)

mAP:

0.1990(FLOAT)/0.1990(INT8)

Nuscenes
Bev_ipm_4d_efficientnetb0_multitask

image:

6x3x512x960

points:

6x128x128x2

prev_feat:

1x164x28x128

prev_point:

1x128x128x2

NDS:

0.3722(FLOAT)/0.3730(INT8)

MeanIOU:

0.5287(FLOAT)/0.5388(INT8)

mAP:

0.2201(FLOAT)/0.2215(INT8)

Nuscenes
Bev_ipm_efficientnetb0_multitask

image:

6x3x512x960

points:

6x128x128x2

NDS:

0.3055(FLOAT)/0.3041(INT8)

MeanIOU:

0.5145(FLOAT)/0.5105(INT8)

mAP:

0.2170(FLOAT)/0.2163(INT8)

Nuscenes
Bev_lss_efficientnetb0_multitask

image:

6x3x256x704

points(0&1):

10x128x128x2

NDS:

0.3006(FLOAT)/0.3003(INT8)

MeanIOU:

0.5180(FLOAT)/0.5171(INT8)

mAP:

0.2061(FLOAT)/0.2046(INT8)

Nuscenes
Detr3d_efficientnetb3

coords(0-3):

6x4x256x2

image:

6x3x512x1408

masks:

1x4x256x24

NDS:

0.3304(FLOAT)/0.3309(INT8)

mAP:

0.2753(FLOAT)/0.2743(INT8)

Nuscenes
Petr_efficientnetb3

image:

6x3x512x1408

pos_embed:

1x96x44x256

NDS:

0.3765(FLOAT)/0.3745(INT8)

mAP:

0.3038(FLOAT)/0.2937(INT8)

Nuscenes
Bevformer_tiny_resnet50_detection

img:

6x3x480x800

prev_bev:

1x2500x256

prev_bev_ref:

1x50x50x2

queries_rebatch_grid:

6x20x32x2

restore_bev_grid:

1x100x50x2

reference_points_rebatch:

6x640x4x2

bev_pillar_counts:

1x2500x1

NDS:

0.3713(FLOAT)/0.3700(INT8)

mAP:

0.2673(FLOAT)/0.2646(INT8)

Nuscenes
Flashocc_henet_lss_occ3d_nuscenes

img:

6x3x512x960

points:

10x128x128x2

points_depth:

10x128x128x2

mIoU:

0.3675(FLOAT)/0.3687(INT8)

Nuscenes
Horizon_swin_transformer

1x3x224x224

Top1:

0.8024(FLOAT)/0.7988(INT8)

ImageNet
Vargnetv2

1x3x224x224

Top1:

0.7342(FLOAT)/0.7328(INT8)

ImageNet
Vit_small

1x3x224x224

Top1:

0.7950(FLOAT)/0.7921(INT8)

ImageNet
Centerpoint_pointpillar

points:

300000x5

voxel_feature:

1x5x20x40000

coors:

40000x4

NDS:

0.5832(FLOAT)/0.5816(INT8)

mAP:

0.4804(FLOAT)/0.4784(INT8)

Nuscenes
Detr_efficientnetb3

1x3x800x1333

[IoU=0.50:0.95]=

0.3720(FLOAT)/0.3604(INT8)

MS COCO
Detr_resnet50

1x3x800x1333

[IoU=0.50:0.95]=

0.3569(FLOAT)/0.3152(INT8)

MS COCO
FCOS3D_efficientnetb0

1x3x512x896

NDS:

0.3061(FLOAT)/0.3030(INT8)

mAP:

0.2133(FLOAT)/0.2069(INT8)

nuscenes
Fcos_efficientnetb0

1x3x512x512

[IoU=0.50:0.95]=

0.3626(FLOAT)/0.3551(INT8)

MS COCO
Ganet_mixvargenet

1x3x320x800

F1Score:

0.7949(FLOAT)/0.7883(INT8)

CuLane
Deformable_detr_resnet50

1x3x800x1333

[IoU=0.50:0.95]=

0.4413(FLOAT)/0.4194(INT8)

MS COCO
Stereonetplus_mixvargenet

2x3x544x960

EPE:

1.1270(FLOAT)/1.1341(INT8)

SceneFlow
Centerpoint_mixvargnet_multitask

points:

300000x5

voxel_feature:

1x5x20x40000

coors:

40000x4

NDS:

0.5809(FLOAT)/0.5753(INT8)

MeanIOU:

0.9128(FLOAT)/0.9121(INT8)

mAP:

0.4727(FLOAT)/0.4626(INT8)

Nuscenes
Unet_mobilenetv1

1x3x1024x2048

mIoU:

0.6801(FLOAT)/0.6767(INT8)

Cityscapes
Densetnt_vectornet

goals_2d:

30x1x2048x2

goals_2d_mask:

30x1x2048x1

instance_mask:

30x1x96x1

lane_feat:

30x9x64x11

traj_feat:

30x19x32x9

minFDA:

1.2975(FLOAT)/1.3021(INT8)

Argoverse 1
Maptroe_henet_tinym_bevformer

img:

6x3x480x800

osm_mask:

1x1x50x100

queries_rebatch_grid:

6x20x100x2

restore_bev_grid:

1x100x100x2

reference_points_rebatch:

6x2000x4x2

bev_pillar_counts:

1x5000x1

mAP:

0.6633(FLOAT)/0.6564(INT8)

Nuscenes
Qcnet_oe

valid_mask:

1x30x10

valid_mask_a2a:

1x10x30x30

agent_type:

1x30x1

x_a_cur:

1x1x30x1,1x1x30x1,1x1x30x1,1x1x30x1

r_pl2a_cur:

1x1x30x80,1x1x30x80,1x1x30x80

r_t_cur:

1x1x30x6,1x1x30x6,1x1x30x6,1x1x30x6

r_a2a_cur:

1x1x30x30,1x1x30x30,1x1x30x30

x_a_mid_emb:

1x30x2x128

x_a:

1x30x6x128

pl_type,is_intersection:

1x80

r_pl2pl:

1x1x80x80,1x1x80x80,1x1x80x80

r_pt2pl:

1x1x80x50,1x1x80x50,1x1x80x50

mask_pl2pl:

1x80x80

magnitude,pt_type,side,mask:

1x80x50

mask_a2m:

1x30x30

mask_dst:

1x30x1

type_pl2pl:

1x80x80

hitrate:

0.8026(FLOAT)/0.7972(INT8)

Argoverse 2

模型性能

MODEL NAMEINPUT SIZELatency(ms)FPSFPS Configuration
ResNet50

1x3x224x224

1.1321153.55thread_num:4
GoogleNet

1x3x224x224

0.6112858.94thread_num:4
EfficientNet_Lite1

1x240x240x3

0.5953225.66thread_num:4
EfficientNet_Lite2

1x260x260x3

0.6932479.25thread_num:4
EfficientNet_Lite3

1x280x280x3

0.8121899.15thread_num:4
EfficientNet_Lite4

1x300x300x3

1.0581302.93thread_num:4
Vargconvnet

1x3x224x224

0.9531454.34thread_num:4
Efficientnasnet_m

1x3x300x300

0.9541455.45thread_num:4
Efficientnasnet_s

1x3x280x280

0.5743210.04thread_num:4
ResNet18

1x3x224x224

0.6482565.83thread_num:4
YOLOv2_Darknet19

1x3x608x608

4.676227.57thread_num:4
YOLOv3_Darknet53

1x3x416x416

5.034210.93thread_num:4
YOLOv5x_v2.0

1x3x672x672

16.35462.35thread_num:4
Centernet_resnet101

1x3x512x512

5.654186.60thread_num:4
YOLOv3_VargDarknet

1x3x416x416

3.568305.23thread_num:4
Deeplabv3plus_efficientnetb0

1x3x1024x2048

6.920150.82thread_num:4
Fastscnn_efficientnetb0

1x3x1024x2048

4.343246.34thread_num:4
Deeplabv3plus_efficientnetm1

1x3x1024x2048

11.23391.48thread_num:4
Deeplabv3plus_efficientnetm2

1x3x1024x2048

15.82964.43thread_num:4
Bev_gkt_mixvargenet_multitask

image:

6x3x512x960

points(0-8):

6x64x64x2

15.07268.78thread_num:4
Bev_ipm_4d_efficientnetb0_multitask

image:

6x3x512x960

points:

6x128x128x2

prev_feat:

1x164x28x128

prev_point:

1x128x128x2

9.468111.83thread_num:4
Bev_ipm_efficientnetb0_multitask

image:

6x3x512x960

points:

6x128x128x2

9.212114.94thread_num:4
Bev_lss_efficientnetb0_multitask

image:

6x3x256x704

points(0&1):

10x128x128x2

5.865186.15thread_num:4
Detr3d_efficientnetb3

coords(0-3):

6x4x256x2

image:

6x3x512x1408

masks:

1x4x256x24

31.17832.51thread_num:4
Petr_efficientnetb3

image:

6x3x512x1408

pos_embed:

1x96x44x256

48.37020.86thread_num:4
Bevformer_tiny_resnet50_detection

img:

6x3x480x800

prev_bev:

1x2500x256

prev_bev_ref:

1x50x50x2

queries_rebatch_grid:

6x20x32x2

restore_bev_grid:

1x100x50x2

reference_points_rebatch:

6x640x4x2

bev_pillar_counts:

1x2500x1

32.14231.54thread_num:4
Flashocc_henet_lss_occ3d_nuscenes

img:

6x3x512x960

points:

10x128x128x2

points_depth:

10x128x128x2

11.13292.86thread_num:6
Horizon_swin_transformer

1x3x224x224

3.391321.81thread_num:4
Vargnetv2

1x3x224x224

0.4944384.88thread_num:4
Vit_small

1x3x224x224

1.955596.83thread_num:4
Centerpoint_pointpillar

points:

300000x5

voxel_feature:

1x5x20x40000

coors:

40000x4

13.048121.67thread_num:4
Detr_efficientnetb3

1x3x800x1333

18.90253.79thread_num:4
Detr_resnet50

1x3x800x1333

24.54441.29thread_num:4
FCOS3D_efficientnetb0

1x3x512x896

2.691450.63thread_num:4
Fcos_efficientnetb0

1x3x512x512

1.3261105.46thread_num:4
Ganet_mixvargenet

1x3x320x800

0.9191578.04thread_num:4
Deformable_detr_resnet50

1x3x800x1333

200.4075.00thread_num:6
Stereonetplus_mixvargenet

2x3x544x960

4.766223.98thread_num:4
Centerpoint_mixvargnet_multitask

points:

300000x5

voxel_feature:

1x5x20x40000

coors:

40000x4

10.370185.18thread_num:4
Unet_mobilenetv1

1x3x1024x2048

1.611812.13thread_num:4
Densetnt_vectornet

goals_2d:

30x1x2048x2

goals_2d_mask:

30x1x2048x1

instance_mask:

30x1x96x1

lane_feat:

30x9x64x11

traj_feat:

30x19x32x9

6.927156.09thread_num:4
Maptroe_henet_tinym_bevformer

img:

6x3x480x800

osm_mask:

1x1x50x100

queries_rebatch_grid:

6x20x100x2

restore_bev_grid:

1x100x100x2

reference_points_rebatch:

6x2000x4x2

bev_pillar_counts:

1x5000x1

13.85674.28thread_num:4
Qcnet_oe

valid_mask:

1x30x10

valid_mask_a2a:

1x10x30x30

agent_type:

1x30x1

x_a_cur:

1x1x30x1,1x1x30x1,1x1x30x1,1x1x30x1

r_pl2a_cur:

1x1x30x80,1x1x30x80,1x1x30x80

r_t_cur:

1x1x30x6,1x1x30x6,1x1x30x6,1x1x30x6

r_a2a_cur:

1x1x30x30,1x1x30x30,1x1x30x30

x_a_mid_emb:

1x30x2x128

x_a:

1x30x6x128

pl_type,is_intersection:

1x80

r_pl2pl:

1x1x80x80,1x1x80x80,1x1x80x80

r_pt2pl:

1x1x80x50,1x1x80x50,1x1x80x50

mask_pl2pl:

1x80x80

magnitude,pt_type,side,mask:

1x80x50

mask_a2m:

1x30x30

mask_dst:

1x30x1

type_pl2pl:

1x80x80

5.094220.29thread_num:4