Model benchmark

Test environment

  • PyTorch 2.2.2

  • Lightning 1.9.5

  • NVIDIA RTX 3090

  • CUDA 12.1

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

382.907

0.708

572.08

1.188

ccmr+

11.524

11716.13

500

1000

500000

855.701

1.64

1386.917

3.026

craft

6.307

4204.715

500

1000

500000

315.168

4.437

475.565

8.178

csflow

5.605

1826.398

500

1000

500000

102.378

3.461

155.09

1.89

dicl

11.226

453.293

500

1000

500000

49.855

1.796

57.079

2.556

dip

5.372

5780.579

500

1000

500000

341.781

1.056

513.528

1.631

fastflownet

1.366

23.72

500

1000

500000

35.841

0.481

29.944

0.514

flow1d

5.734

1523.984

500

1000

500000

86.135

0.745

122.418

0.764

flowformer

16.168

3207.109

500

1000

500000

220.047

3.186

371.532

6.111

flowformer++

16.152

3048.144

500

1000

500000

212.087

2.895

356.207

4.428

flownet2

162.519

358.356

500

1000

500000

58.25

1.338

69.212

1.847

flownetc

39.175

82.78

500

1000

500000

27.886

1.157

30.459

0.801

flownetcs

77.871

139.998

500

1000

500000

35.427

0.827

39.28

0.953

flownetcss

116.566

197.216

500

1000

500000

42.829

0.991

48.338

1.096

flownets

38.677

53.172

500

1000

500000

4.851

0.876

5.652

0.69

flownetsd

45.372

95.366

500

1000

500000

7.126

0.77

8.049

0.889

gma

5.88

2540.319

500

1000

500000

106.156

1.086

189.866

1.438

gmflow

4.68

428.6

500

1000

500000

28.921

0.894

50.696

1.402

gmflow_refine

4.717

1063.111

500

1000

500000

72.332

1.663

132.252

2.987

gmflow+

4.68

428.6

500

1000

500000

28.98

0.887

52.061

1.401

gmflow+_sc2

4.717

1063.111

500

1000

500000

74.425

1.663

135.823

2.977

gmflow+_sc2_refine6

7.361

2110.617

500

1000

500000

145.88

1.669

244.301

2.989

gmflownet

9.343

2059.582

500

1000

500000

151.625

1.512

246.5

2.52

gmflownet_mix

8.688

1871.501

500

1000

500000

127.058

1.237

198.365

2.038

hd3

39.562

289.931

500

1000

500000

53.431

0.935

60.143

1.213

hd3_ctxt

39.943

310.383

500

1000

500000

53.431

0.98

56.934

1.223

irr_pwc

6.362

907.539

500

1000

500000

130.299

0.919

157.042

1.303

irr_pwcnet

8.639

178.394

500

1000

500000

42.655

0.725

43.146

0.989

irr_pwcnet_irr

3.354

170.375

500

1000

500000

41.807

0.725

49.523

0.946

lcv_raft

5.323

1781.566

500

1000

500000

162.396

1.252

lcv_raft_small

1.007

459.632

500

1000

500000

88.596

1.219

liteflownet

5.38

310.077

500

1000

500000

65.545

0.729

74.459

0.914

liteflownet2

6.429

140.855

500

1000

500000

46.831

0.733

39.062

0.799

liteflownet2_pseudoreg

6.493

148.229

500

1000

500000

37.606

0.735

41.947

0.818

liteflownet3

7.524

179.333

500

1000

500000

62.753

0.749

61.946

0.812

liteflownet3_pseudoreg

7.588

186.706

500

1000

500000

61.368

0.748

65.64

0.812

liteflownet3s

8.006

181.128

500

1000

500000

69.932

0.745

68.469

0.816

liteflownet3s_pseudoreg

8.07

188.501

500

1000

500000

74.508

0.747

69.647

0.816

llaflow

6.06

3313.444

500

1000

500000

125.561

1.461

244.039

2.512

llaflow_raft

5.52

2553.575

500

1000

500000

115.749

1.49

198.493

2.672

maskflownet

20.656

315.189

500

1000

500000

76.374

0.745

92.676

0.759

maskflownet_s

10.514

162.96

500

1000

500000

47.269

0.733

62.912

1.057

memflow

6.273

1628.784

500

1000

500000

157.404

1.156

244.581

1.861

memflow_t

12.705

1581.625

500

1000

500000

159.091

1.313

259.675

2.267

ms_raft+

16.177

9658.707

500

1000

500000

525.469

1.477

756.039

2.775

neuflow

3.847

76.246

500

1000

500000

14.884

0.668

15.324

0.673

neuflow2

9.03

255.868

500

1000

500000

22.123

0.903

20.515

1.125

pwcnet

9.374

157.505

500

1000

500000

48.76

0.713

46.366

1.072

pwcnet_nodc

8.243

96.719

500

1000

500000

44.44

0.666

43.733

0.986

raft

5.258

1780.45

500

1000

500000

124.247

0.916

145.654

1.254

raft_small

0.99

459.365

500

1000

500000

88.98

0.922

88.807

1.213

rapidflow

1.646

114.735

500

1000

500000

22.244

0.622

30.773

1.007

rapidflow_it1

1.646

12.155

500

1000

500000

6.7

0.512

8.722

0.58

rapidflow_it2

1.646

17.449

500

1000

500000

9.363

0.513

11.398

0.585

rapidflow_it3

1.646

51.45

500

1000

500000

11.617

0.628

16.119

1.009

rapidflow_it6

1.646

72.545

500

1000

500000

15.331

0.616

20.98

1.005

rpknet

2.847

264.55

500

1000

500000

112.008

0.686

148.554

1.118

sea_raft

8.883

538.417

500

1000

500000

29.095

0.854

45.788

1.203

sea_raft_s

8.883

538.417

500

1000

500000

29.475

0.856

46.55

1.203

sea_raft_m

19.664

922.464

500

1000

500000

40.657

0.893

67.91

1.234

sea_raft_l

19.664

1242.004

500

1000

500000

66.027

0.878

105.991

1.22

scopeflow

6.362

907.539

500

1000

500000

180.012

0.726

198.476

1.202

separableflow

8.346

1658.256

500

1000

500000

602.443

1.286

skflow

6.273

2731.596

500

1000

500000

290.073

1.013

362.867

1.467

starflow

4.772

786.3

500

1000

500000

177.326

0.93

189.626

1.314

unimatch

4.68

428.6

500

1000

500000

27.738

0.92

52.008

1.421

unimatch_sc2

4.717

1063.111

500

1000

500000

75.261

1.684

136.203

3.006

unimatch_sc2_refine6

7.361

2110.617

500

1000

500000

145.863

1.689

243.728

3.02

vcn

10.311

86.289

500

1000

500000

278.101

1.291

424.59

2.107

vcn_small

8.371

29.129

500

1000

500000

118.486

0.782

161.606

1.001

videoflow_bof

12.659

3159.379

500

1000

500000

420.908

1.453

481.482

2.456

videoflow_mof

13.453

3460.933

500

1000

500000

471.891

1.439

545.409

2.464