Model benchmark

Test environment

  • PyTorch 2.1.0

  • Lightning 1.9.5

  • NVIDIA RTX 3070

  • CUDA 11.7

Model

Params

FLOPs

InputH

InputW

InputPx

Time(ms)-fp16

Memory(GB)-fp16

Time(ms)-fp32

Memory(GB)-fp32

ccmr

10.781

4134.022

500

1000

500000

509.844

0.971

789.358

1.452

ccmr+

11.524

11716.13

500

1000

500000

1185.023

1.899

1990.497

3.131

craft

6.307

4204.715

500

1000

500000

428.948

4.633

csflow

5.605

1826.398

500

1000

500000

147.029

3.25

266.61

2.176

dicl

11.226

453.293

500

1000

500000

52.43

1.934

73.607

2.782

dip

5.372

5780.579

500

1000

500000

490.152

1.27

859.917

1.784

fastflownet

1.366

23.72

500

1000

500000

27.487

0.698

31.928

0.731

flow1d

5.734

1523.984

500

1000

500000

117.403

0.961

206.983

1.061

flowformer

16.168

3207.073

500

1000

500000

312.59

3.428

575.758

6.327

flowformer++

16.152

3048.107

500

1000

500000

306.335

2.655

541.38

4.643

flownet2

162.519

358.356

500

1000

500000

64.286

1.553

109.808

2.141

flownetc

39.175

82.78

500

1000

500000

27.385

1.366

47.936

1.063

flownetcs

77.871

139.998

500

1000

500000

34.673

1.042

60.489

1.215

flownetcss

116.566

197.216

500

1000

500000

41.98

1.204

72.73

1.356

flownets

38.677

53.172

500

1000

500000

5.686

1.071

10.296

0.938

flownetsd

45.372

95.366

500

1000

500000

8.738

0.922

13.76

1.092

gma

5.88

2540.319

500

1000

500000

178.967

1.29

316.078

1.729

gmflow

4.68

428.6

500

1000

500000

46.871

1.098

86.475

1.68

gmflow_refine

4.717

1063.111

500

1000

500000

116.638

1.893

214.12

3.266

gmflow+

4.68

428.6

500

1000

500000

47.648

1.086

87.964

1.68

gmflow+_sc2

4.717

1063.111

500

1000

500000

120.369

1.893

221.429

3.266

gmflow+_sc2_refine6

7.361

2110.617

500

1000

500000

219.626

1.899

389.602

3.284

gmflownet

9.343

2059.582

500

1000

500000

212.103

1.725

400.303

2.739

gmflownet_mix

8.688

1871.501

500

1000

500000

178.81

1.45

326.366

2.254

hd3

39.562

289.931

500

1000

500000

55.854

1.176

70.501

1.422

hd3_ctxt

39.943

310.383

500

1000

500000

56.623

1.182

74.286

1.432

irr_pwc

6.362

907.539

500

1000

500000

135.557

1.161

184.908

1.305

irr_pwcnet

8.639

178.394

500

1000

500000

41.291

0.938

52.19

1.278

irr_pwcnet_irr

3.354

170.375

500

1000

500000

39.302

0.866

49.308

1.112

lcv_raft

5.323

1781.566

500

1000

500000

258.342

1.543

lcv_raft_small

1.007

459.632

500

1000

500000

99.857

1.428

liteflownet

5.38

310.077

500

1000

500000

64.352

0.948

82.841

1.145

liteflownet2

6.429

140.855

500

1000

500000

38.297

0.807

42.396

0.883

liteflownet2_pseudoreg

6.493

148.229

500

1000

500000

35.59

0.809

44.021

0.883

liteflownet3

7.524

179.333

500

1000

500000

55.668

0.938

70.084

1.012

liteflownet3_pseudoreg

7.588

186.706

500

1000

500000

53.77

0.938

74.125

1.012

liteflownet3s

8.006

181.128

500

1000

500000

63.587

0.938

69.005

1.014

liteflownet3s_pseudoreg

8.07

188.501

500

1000

500000

56.794

0.94

73.9

1.014

llaflow

6.06

3313.444

500

1000

500000

208.742

1.649

391.392

2.768

llaflow_raft

5.52

2553.575

500

1000

500000

169.808

1.52

330.792

2.917

maskflownet

20.656

315.189

500

1000

500000

70.273

0.946

90.739

1.284

maskflownet_s

10.514

162.96

500

1000

500000

40.171

0.936

50.655

1.2

matchflow

15.446

2994.565

500

1000

500000

440.889

2.196

matchflow_raft

14.824

2284.576

500

1000

500000

373.69

2.022

ms_raft+

16.177

9658.707

500

1000

500000

795.76

1.709

1197.577

2.965

neuflow

3.847

76.246

500

1000

500000

9.22

0.52

13.435

0.508

pwcnet

9.374

157.505

500

1000

500000

39.595

0.917

47.103

1.209

pwcnet_nodc

8.243

96.719

500

1000

500000

36.401

0.874

41.83

1.147

raft

5.258

1780.45

500

1000

500000

140.326

1.233

256.396

1.543

raft_small

0.99

459.365

500

1000

500000

56.81

1.125

87.687

1.418

rapidflow

1.646

99.618

500

1000

500000

38.527

0.85

45.661

1.008

rapidflow_it1

1.646

12.099

500

1000

500000

8.038

0.842

8.345

0.877

rapidflow_it2

1.646

16.507

500

1000

500000

10.596

0.846

11.207

0.883

rapidflow_it3

1.646

36.336

500

1000

500000

15.68

0.848

17.682

1.008

rapidflow_it6

1.646

57.43

500

1000

500000

23.137

0.848

26.834

1.008

rpknet

2.847

249.432

500

1000

500000

169.666

0.917

182.703

1.104

scopeflow

6.362

907.539

500

1000

500000

157.852

0.93

194.548

1.219

separableflow

8.346

1658.256

500

1000

500000

672.329

1.579

skflow

6.273

2731.596

500

1000

500000

367.02

1.354

578.761

1.77

starflow

4.772

786.3

500

1000

500000

134.952

1.149

180.197

1.592

vcn

10.311

86.289

500

1000

500000

263.144

1.497

396.497

2.333

vcn_small

8.371

29.129

500

1000

500000

95.822

0.938

141.674

1.278

videoflow_bof

12.659

3159.33

500

1000

500000

473.373

1.809

747.154

2.672

videoflow_mof

13.453

3460.884

500

1000

500000

528.998

1.795

819.146

2.684