Edit model card

segformer-b0-finetuned-segments-ECHO-dev-05-v1

This model is a fine-tuned version of nvidia/mit-b0 on the unreal-hug/REAL_DATASET_SEG dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4592
  • Mean Iou: 0.3826
  • Mean Accuracy: 0.5892
  • Overall Accuracy: 0.5467
  • Accuracy Unlabeled: nan
  • Accuracy Lv: 0.7143
  • Accuracy Lr: 0.4323
  • Accuracy Ra: 0.7629
  • Accuracy La: 0.4472
  • Iou Unlabeled: 0.0
  • Iou Lv: 0.7065
  • Iou Lr: 0.4317
  • Iou Ra: 0.4223
  • Iou La: 0.3527

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: 6e-05
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 50

Training results

Training Loss Epoch Step Validation Loss Mean Iou Mean Accuracy Overall Accuracy Accuracy Unlabeled Accuracy Lv Accuracy Lr Accuracy Ra Accuracy La Iou Unlabeled Iou Lv Iou Lr Iou Ra Iou La
1.1252 2.86 20 1.3259 0.1971 0.3379 0.4375 nan 0.0 0.6365 0.5291 0.1860 0.0 0.0 0.4923 0.3492 0.1439
0.9104 5.71 40 0.9589 0.1818 0.3421 0.3596 nan 0.0145 0.3590 0.7644 0.2304 0.0 0.0144 0.3436 0.3778 0.1731
0.7567 8.57 60 0.7761 0.2203 0.3739 0.3852 nan 0.0808 0.3882 0.6422 0.3844 0.0 0.0803 0.3778 0.4073 0.2360
0.7035 11.43 80 0.7442 0.2729 0.4718 0.4941 nan 0.2145 0.5077 0.8370 0.3279 0.0 0.2134 0.4817 0.4073 0.2619
0.5781 14.29 100 0.6260 0.2876 0.4446 0.4279 nan 0.4235 0.3777 0.5787 0.3986 0.0 0.3683 0.3761 0.4063 0.2873
0.5438 17.14 120 0.5559 0.3877 0.5412 0.5761 nan 0.5803 0.6504 0.5190 0.4149 0.0 0.5671 0.6193 0.4171 0.3352
0.5198 20.0 140 0.5617 0.3724 0.5617 0.5335 nan 0.6661 0.4532 0.7059 0.4216 0.0 0.6419 0.4532 0.4129 0.3540
0.4435 22.86 160 0.5393 0.4160 0.6198 0.6126 nan 0.7555 0.5832 0.6962 0.4442 0.0 0.7000 0.5705 0.4873 0.3221
0.5002 25.71 180 0.5126 0.4094 0.6080 0.6043 nan 0.6854 0.5833 0.6945 0.4687 0.0 0.6771 0.5761 0.4762 0.3176
0.4142 28.57 200 0.4874 0.3503 0.5361 0.4949 nan 0.6967 0.3895 0.6436 0.4147 0.0 0.6287 0.3895 0.4106 0.3228
0.3092 31.43 220 0.4819 0.3857 0.6001 0.5534 nan 0.7296 0.4267 0.8020 0.4423 0.0 0.7157 0.4267 0.4266 0.3595
0.2895 34.29 240 0.4969 0.3983 0.6220 0.5809 nan 0.7353 0.4689 0.8050 0.4787 0.0 0.7265 0.4677 0.4474 0.3498
0.3046 37.14 260 0.4767 0.4248 0.6412 0.6115 nan 0.7853 0.5270 0.7814 0.4711 0.0 0.7712 0.5199 0.4587 0.3742
0.3514 40.0 280 0.4531 0.3978 0.5989 0.5767 nan 0.7112 0.5082 0.7478 0.4282 0.0 0.6979 0.5024 0.4353 0.3537
0.2891 42.86 300 0.4629 0.3842 0.5885 0.5488 nan 0.7046 0.4397 0.7693 0.4403 0.0 0.6982 0.4366 0.4237 0.3623
0.2512 45.71 320 0.4584 0.3783 0.5794 0.5357 nan 0.7144 0.4199 0.7390 0.4443 0.0 0.7016 0.4196 0.4134 0.3568
0.2695 48.57 340 0.4592 0.3826 0.5892 0.5467 nan 0.7143 0.4323 0.7629 0.4472 0.0 0.7065 0.4317 0.4223 0.3527

Framework versions

  • Transformers 4.35.2
  • Pytorch 2.1.0+cu118
  • Datasets 2.15.0
  • Tokenizers 0.15.0
Downloads last month
2
Safetensors
Model size
3.72M params
Tensor type
F32
·

Finetuned from