CoolFace
Modelpublic

woodBorjo/finetune-instance-segmentation-ade20k-mini-mask2former

sourceHugging Faceotherupdated 11mo agoView on Hugging Face
0likes11downloads
Model Card

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

finetune-instance-segmentation-ade20k-mini-mask2former

This model is a fine-tuned version of facebook/mask2former-swin-tiny-coco-instance on the qubvel-hf/ade20k-mini dataset. It achieves the following results on the evaluation set:

  • —Loss: 28.6625
  • —Map: 0.2266
  • —Map 50: 0.4359
  • —Map 75: 0.2107
  • —Map Small: 0.1469
  • —Map Medium: 0.6658
  • —Map Large: 0.8156
  • —Mar 1: 0.0959
  • —Mar 10: 0.2562
  • —Mar 100: 0.2914
  • —Mar Small: 0.2182
  • —Mar Medium: 0.7126
  • —Mar Large: 0.8476
  • —Map Person: 0.159
  • —Mar 100 Person: 0.2146
  • —Map Car: 0.2941
  • —Mar 100 Car: 0.3683

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • —learning_rate: 1e-05
  • —trainbatchsize: 8
  • —evalbatchsize: 8
  • —seed: 42
  • —gradientaccumulationsteps: 2
  • —totaltrainbatch_size: 16
  • —optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • —lrschedulertype: constant
  • —num_epochs: 40.0
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossMapMap 50Map 75Map SmallMap MediumMap LargeMar 1Mar 10Mar 100Mar SmallMar MediumMar LargeMap PersonMar 100 PersonMap CarMar 100 Car
34.53691.010032.52210.16980.34330.15140.1080.58880.79270.08810.23690.28070.20630.70720.86250.12440.19820.21510.3632
28.44762.020030.46100.19270.38940.17410.12650.61250.81540.09380.24510.28450.21070.70760.85940.1340.20150.25130.3675
26.90793.030029.23910.20240.40590.18740.13580.63130.82120.09420.25210.28910.21550.7120.85940.13770.20550.26710.3728
25.91744.040028.71600.21080.41570.19690.14030.63790.80890.09570.25240.28730.21410.71040.83160.14470.2070.2770.3677
25.34215.050028.39160.21470.42240.20030.14210.64610.83210.09610.25270.29180.21850.71190.86490.14830.20930.28110.3743
24.71766.060028.29110.21850.42260.20640.14310.65280.83010.09710.25720.2930.21970.71380.85620.14990.21180.28710.3741
24.137.070027.77880.220.42990.2070.14510.6490.7990.09720.25880.29330.2210.71070.83160.1540.21270.2860.374
23.71078.080027.75780.21760.42470.20270.14230.6540.83460.09530.25530.29350.21990.71630.86490.15220.21190.28310.3752
23.42649.090027.44610.22080.42890.21140.1430.65390.830.09610.25450.29530.2220.71540.86490.1530.21160.28860.3789
23.226210.0100027.75150.22380.43620.21120.14680.65320.83060.09560.25670.29440.22130.71360.86180.15410.21010.29360.3786
22.874311.0110027.94570.2260.4420.21480.14950.65190.80960.09670.25550.29310.2210.70860.83650.15550.21220.29640.3741
22.499512.0120027.74850.22610.43640.2170.14780.65820.83320.09730.2560.29590.22330.71280.85940.15750.21370.29470.378
22.430313.0130027.81370.22330.42660.21620.14590.65510.83310.09710.25380.29030.21660.71310.85870.15520.21120.29140.3693
22.167314.0140027.45360.22530.43560.21760.14790.65540.82830.09840.25470.29310.22030.71160.85310.15640.21480.29430.3714
21.827215.0150027.36260.22540.43980.21530.14750.65790.79870.09770.25320.29220.22010.70920.82290.15720.21180.29370.3726
21.644316.0160027.13250.22860.44520.2170.15020.65860.81470.09880.25780.29610.22350.7140.84380.15630.21460.3010.3776
21.532617.0170027.59650.22750.440.21920.14820.65940.82210.09860.25840.29540.22270.71340.85310.15840.21440.29660.3764
21.47218.0180027.73450.22820.43680.21460.150.65770.82170.09830.25590.29580.2230.7140.85310.15710.21350.29920.378
21.05319.0190027.46630.22660.43830.21430.1480.65690.81770.09770.25480.29050.21770.70970.84130.15710.20950.2960.3714
20.917920.0200027.50200.2260.4410.21150.14790.66010.79010.09760.25460.29270.22030.71150.81670.15640.21350.29550.3719
20.696921.0210027.25020.22960.44210.2140.1510.66190.82990.09920.2570.29380.2210.7120.85310.15860.21390.30070.3737
20.541922.0220027.74220.23070.44560.21690.15330.65680.82840.09950.25750.29460.22210.71130.85310.15810.21320.30330.3761
20.533423.0230027.05270.22710.43770.21710.14820.6630.80930.09660.25420.29040.21720.71250.84130.15590.20920.29830.3717
20.324824.0240027.81910.22990.44350.21550.15180.65720.820.09910.25770.29570.22370.70860.85310.15710.21130.30260.38
20.235425.0250027.68900.22830.44230.21540.15060.65920.82930.09840.25690.29640.22380.71350.85620.15650.21460.30.3782
20.133626.0260028.01790.22820.43460.21320.14990.66480.8130.09760.25530.29280.21910.71750.84510.15590.21390.30040.3717
19.761527.0270027.93830.22870.44210.2160.15120.65940.81050.09950.25720.2930.220.71310.84440.15820.21480.29910.3711
19.783328.0280027.56690.22830.44030.21460.15010.65670.79390.09760.25730.29220.22040.70790.81670.15730.21370.29930.3708
19.669629.0290027.63340.22830.43570.21570.1490.65960.78610.09770.25460.28950.21680.71080.81420.15670.21060.29980.3684
19.517730.0300028.15260.23190.44290.21350.15280.66260.82030.09730.25770.29570.22280.71460.85240.15930.21350.30450.3779
19.480631.0310027.88460.2270.43980.21430.14990.65640.82110.09650.25640.29160.21860.7120.84380.15840.21380.29570.3695
19.19432.0320028.02000.22860.44070.21610.15010.66160.80880.09710.2560.29110.21790.71340.83260.16030.21470.29690.3675
19.227833.0330027.32230.23120.43850.21680.15110.66260.80970.09790.25470.29160.21910.7090.8420.16040.2120.3020.3711
19.04834.0340028.11980.22840.44170.21480.15070.66030.81970.09810.25660.29330.22060.71140.850.1580.21450.29880.3722
18.900235.0350028.72060.22990.43790.21550.15240.66010.81730.09830.25970.2940.22140.71180.84760.16010.21320.29960.3747
18.944536.0360028.21010.23010.44420.21440.15060.66130.82140.09830.25770.29310.22050.71020.85070.15990.21420.30040.3719
18.708737.0370028.82120.22990.44370.21130.150.66980.79720.09870.25740.29270.21920.71880.82360.16210.2150.29780.3704
18.748738.0380028.73130.22930.44050.21290.14890.66490.79810.09750.25760.29410.22160.71340.82530.15980.21470.29880.3735
18.59439.0390028.36690.23040.44310.21380.15220.66190.80480.09740.25920.29730.22520.71170.850.16120.21460.29950.38
18.446240.0400028.66250.22660.43590.21070.14690.66580.81560.09590.25620.29140.21820.71260.84760.1590.21460.29410.3683

Framework versions

  • —Transformers 4.57.1
  • —Pytorch 2.9.0+cu128
  • —Datasets 4.3.0
  • —Tokenizers 0.22.1