unreal-hug's picture
End of training
579a8a6 verified
metadata
license: other
base_model: nvidia/mit-b5
tags:
  - vision
  - image-segmentation
  - generated_from_trainer
model-index:
  - name: segformer-b5-seed42-outputs
    results: []

segformer-b5-seed42-outputs

This model is a fine-tuned version of nvidia/mit-b5 on the unreal-hug/REAL_DATASET_SEG_401_6_lbls dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2833
  • Mean Iou: 0.3430
  • Mean Accuracy: 0.4050
  • Overall Accuracy: 0.6546
  • Accuracy Unlabeled: nan
  • Accuracy Lv: 0.7625
  • Accuracy Rv: 0.6171
  • Accuracy Ra: 0.7072
  • Accuracy La: 0.6623
  • Accuracy Vs: 0.0
  • Accuracy As: 0.0
  • Accuracy Mk: 0.0227
  • Accuracy Tk: nan
  • Accuracy Asd: 0.3003
  • Accuracy Vsd: 0.4268
  • Accuracy Ak: 0.5517
  • Iou Unlabeled: 0.0
  • Iou Lv: 0.7175
  • Iou Rv: 0.5629
  • Iou Ra: 0.6665
  • Iou La: 0.5980
  • Iou Vs: 0.0
  • Iou As: 0.0
  • Iou Mk: 0.0207
  • Iou Tk: nan
  • Iou Asd: 0.2802
  • Iou Vsd: 0.3970
  • Iou Ak: 0.5307

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: 0.0001
  • 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: cosine
  • lr_scheduler_warmup_ratio: 0.05
  • training_steps: 1000

Training results

Training Loss Epoch Step Validation Loss Mean Iou Mean Accuracy Overall Accuracy Accuracy Unlabeled Accuracy Lv Accuracy Rv Accuracy Ra Accuracy La Accuracy Vs Accuracy As Accuracy Mk Accuracy Tk Accuracy Asd Accuracy Vsd Accuracy Ak Iou Unlabeled Iou Lv Iou Rv Iou Ra Iou La Iou Vs Iou As Iou Mk Iou Tk Iou Asd Iou Vsd Iou Ak
0.4921 0.62 100 0.4897 0.0906 0.1245 0.3551 nan 0.7217 0.0098 0.0751 0.4344 0.0 0.0 0.0 nan 0.0 0.0 0.0043 0.0 0.5846 0.0097 0.0746 0.3230 0.0 0.0 0.0 nan 0.0 0.0 0.0043
0.3534 1.25 200 0.3565 0.2420 0.3017 0.5592 nan 0.7947 0.4293 0.5941 0.3492 0.0 0.0 0.0 nan 0.1737 0.1500 0.5262 0.0 0.7343 0.3850 0.3997 0.3289 0.0 0.0 0.0 nan 0.1681 0.1456 0.5007
0.4663 1.88 300 0.3434 0.3620 0.4497 0.7082 nan 0.8026 0.7564 0.6551 0.7392 0.0 0.0 0.0 nan 0.3118 0.5764 0.6560 0.0 0.7545 0.6572 0.6039 0.5921 0.0 0.0 0.0 nan 0.2819 0.4908 0.6019
0.1737 2.5 400 0.3055 0.3331 0.4090 0.6394 nan 0.7469 0.6281 0.5765 0.6122 0.0 0.0 0.0004 nan 0.2401 0.6135 0.6724 0.0 0.7075 0.5704 0.5194 0.5310 0.0 0.0 0.0003 nan 0.2292 0.5279 0.5789
0.1954 3.12 500 0.3052 0.2570 0.2980 0.5174 nan 0.6624 0.4973 0.4223 0.5361 0.0 0.0 0.0022 nan 0.1117 0.3193 0.4284 0.0 0.6289 0.4592 0.4113 0.4902 0.0 0.0 0.0022 nan 0.1107 0.3024 0.4216
0.2666 3.75 600 0.3177 0.3808 0.4720 0.7175 nan 0.7675 0.7191 0.8483 0.7341 0.0 0.0 0.0950 nan 0.3086 0.6065 0.6405 0.0 0.7200 0.6353 0.6912 0.6409 0.0 0.0 0.0845 nan 0.2905 0.5245 0.6024
0.2214 4.38 700 0.2971 0.3748 0.4463 0.7178 nan 0.8524 0.6207 0.7488 0.7353 0.0 0.0 0.025 nan 0.3236 0.5440 0.6130 0.0 0.7909 0.5707 0.6987 0.6564 0.0 0.0 0.0235 nan 0.3015 0.4902 0.5907
0.2624 5.0 800 0.2833 0.3430 0.4050 0.6546 nan 0.7625 0.6171 0.7072 0.6623 0.0 0.0 0.0227 nan 0.3003 0.4268 0.5517 0.0 0.7175 0.5629 0.6665 0.5980 0.0 0.0 0.0207 nan 0.2802 0.3970 0.5307
0.3578 5.62 900 0.2847 0.3329 0.3926 0.6257 nan 0.7276 0.5712 0.6573 0.6410 0.0016 0.0 0.0227 nan 0.3125 0.4450 0.5470 0.0 0.6860 0.5210 0.6234 0.5790 0.0015 0.0 0.0210 nan 0.2906 0.4122 0.5275
0.2736 6.25 1000 0.2861 0.3393 0.4010 0.6425 nan 0.7587 0.5808 0.6702 0.6477 0.0014 0.0 0.0244 nan 0.3087 0.4702 0.5477 0.0 0.7133 0.5292 0.6319 0.5844 0.0014 0.0 0.0225 nan 0.2877 0.4328 0.5295

Framework versions

  • Transformers 4.37.2
  • Pytorch 2.1.2+cu121
  • Datasets 2.16.1
  • Tokenizers 0.15.0