ModelListedExternal

Autocrop Tekst

This model is a fine-tuned version of nvidia/mit-b0 on the /mnt/disk1/autocrop-data/datasets/tekst dataset. It achieves the following results on the evaluation set: Loss: 0.0197 Mean Iou: 0.4970 Mean Accuracy: 0.9939...

Sourcehuggingface.co/NbAiLab/autocrop-tekst/tree/fb1d6422cabb2e288607aebc396cc4f6a5956155

KindOtherBase modelnvidia/mit-b0VersionRevision fb1d642Cgrade

Model source

Kind: Other
File size: 14.2 MB
Base model: nvidia/mit-b0
Version: vRevision fb1d642
Updated: Sep 16, 2026
Downloads: 3
Source: Hugging Face

Source description

autocrop-tekst

This model is a fine-tuned version of nvidia/mit-b0 on the /mnt/disk1/autocrop-data/datasets/tekst dataset. It achieves the following results on the evaluation set:

Loss: 0.0197
Mean Iou: 0.4970
Mean Accuracy: 0.9939
Overall Accuracy: 0.9939
Accuracy Background: nan
Accuracy Crop: 0.9939
Iou Background: 0.0
Iou Crop: 0.9939

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: 8
eval_batch_size: 8
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 0.1
num_epochs: 50.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Crop	Iou Crop
0.1117	1.0	624	0.0847	0.4905	0.9810	0.9810	nan	0.9810	0.9810
0.0532	2.0	1248	0.0444	0.4940	0.9880	0.9880	nan	0.9880	0.9880
0.0382	3.0	1872	0.0330	0.4962	0.9924	0.9924	nan	0.9924	0.9924
0.0302	4.0	2496	0.0292	0.4954	0.9908	0.9908	nan	0.9908	0.9908
0.0294	5.0	3120	0.0279	0.4946	0.9892	0.9892	nan	0.9892	0.9892
0.0284	6.0	3744	0.0255	0.4962	0.9925	0.9925	nan	0.9925	0.9925
0.0256	7.0	4368	0.0249	0.4953	0.9906	0.9906	nan	0.9906	0.9906
0.0294	8.0	4992	0.0242	0.4949	0.9898	0.9898	nan	0.9898	0.9898
0.0262	9.0	5616	0.0238	0.4963	0.9927	0.9927	nan	0.9927	0.9927
0.0262	10.0	6240	0.0230	0.4959	0.9918	0.9918	nan	0.9918	0.9918
0.0306	11.0	6864	0.0225	0.4965	0.9930	0.9930	nan	0.9930	0.9930
0.0223	12.0	7488	0.0221	0.4961	0.9921	0.9921	nan	0.9921	0.9921
0.0231	13.0	8112	0.0215	0.4963	0.9926	0.9926	nan	0.9926	0.9926
0.0208	14.0	8736	0.0215	0.4965	0.9931	0.9931	nan	0.9931	0.9931
0.0203	15.0	9360	0.0215	0.4966	0.9933	0.9933	nan	0.9933	0.9933
0.0222	16.0	9984	0.0211	0.4970	0.9940	0.9940	nan	0.9940	0.9940
0.0230	17.0	10608	0.0211	0.4967	0.9935	0.9935	nan	0.9935	0.9935
0.0228	18.0	11232	0.0211	0.4975	0.9950	0.9950	nan	0.9950	0.9950
0.0211	19.0	11856	0.0216	0.4968	0.9936	0.9936	nan	0.9936	0.9936
0.0201	20.0	12480	0.0211	0.4973	0.9945	0.9945	nan	0.9945	0.9945
0.0198	21.0	13104	0.0204	0.4969	0.9938	0.9938	nan	0.9938	0.9938
0.0204	22.0	13728	0.0206	0.4967	0.9934	0.9934	nan	0.9934	0.9934
0.0192	23.0	14352	0.0201	0.4965	0.9931	0.9931	nan	0.9931	0.9931
0.0201	24.0	14976	0.0202	0.4973	0.9946	0.9946	nan	0.9946	0.9946
0.0192	25.0	15600	0.0202	0.4971	0.9943	0.9943	nan	0.9943	0.9943
0.0189	26.0	16224	0.0197	0.4970	0.9939	0.9939	nan	0.9939	0.9939
0.0184	27.0	16848	0.0200	0.4970	0.9940	0.9940	nan	0.9940	0.9940
0.0179	28.0	17472	0.0201	0.4974	0.9948	0.9948	nan	0.9948	0.9948
0.0192	29.0	18096	0.0198	0.4971	0.9941	0.9941	nan	0.9941	0.9941
0.0176	30.0	18720	0.0199	0.4969	0.9938	0.9938	nan	0.9938	0.9938
0.0175	31.0	19344	0.0200	0.4972	0.9945	0.9945	nan	0.9945	0.9945
0.0158	32.0	19968	0.0200	0.4972	0.9944	0.9944	nan	0.9944	0.9944
0.0173	33.0	20592	0.0204	0.4971	0.9943	0.9943	nan	0.9943	0.9943
0.0172	34.0	21216	0.0201	0.4972	0.9943	0.9943	nan	0.9943	0.9943
0.0164	35.0	21840	0.0201	0.4972	0.9945	0.9945	nan	0.9945	0.9945
0.0156	36.0	22464	0.0199	0.4972	0.9944	0.9944	nan	0.9944	0.9944
0.0164	37.0	23088	0.0202	0.4973	0.9947	0.9947	nan	0.9947	0.9947
0.0159	38.0	23712	0.0201	0.4974	0.9947	0.9947	nan	0.9947	0.9947
0.0160	39.0	24336	0.0200	0.4970	0.9940	0.9940	nan	0.9940	0.9940
0.0169	40.0	24960	0.0201	0.4971	0.9942	0.9942	nan	0.9942	0.9942
0.0152	41.0	25584	0.0201	0.4970	0.9941	0.9941	nan	0.9941	0.9941
0.0159	42.0	26208	0.0201	0.4972	0.9944	0.9944	nan	0.9944	0.9944
0.0177	43.0	26832	0.0200	0.4970	0.9941	0.9941	nan	0.9941	0.9941
0.0142	44.0	27456	0.0201	0.4971	0.9942	0.9942	nan	0.9942	0.9942
0.0176	45.0	28080	0.0201	0.4972	0.9944	0.9944	nan	0.9944	0.9944
0.0159	46.0	28704	0.0203	0.4973	0.9946	0.9946	nan	0.9946	0.9946
0.0149	47.0	29328	0.0202	0.4972	0.9944	0.9944	nan	0.9944	0.9944
0.0159	48.0	29952	0.0201	0.4973	0.9946	0.9946	nan	0.9946	0.9946
0.0142	49.0	30576	0.0202	0.4972	0.9944	0.9944	nan	0.9944	0.9944
0.0139	50.0	31200	0.0201	0.4972	0.9944	0.9944	nan	0.9944	0.9944

Framework versions

Transformers 5.8.0
Pytorch 2.11.0+cu130
Datasets 4.8.5
Tokenizers 0.22.2

Sources

1 source

Hugging FacePrimaryVersion fb1d6422cabb2e288607aebc396cc4f6a5956155

huggingface.co/NbAiLab/autocrop-tekst/tree/fb1d6422cabb2e288607aebc396cc4f6a5956155

Verified Sep 1

Model artifacts

1 artifact

model.safetensors

safetensors · 14.2 MB · SHA-256 3fbb6e68f92e…821a · Hugging Face

Download file

Source excerpts

2 excerpts

ReadmeSep 1huggingface.co/NbAiLab/autocrop-tekst

--- library_name: transformers license: other base_model: nvidia/mit-b0 tags: - image-segmentation - vision - generated_from_trainer datasets: - generator model-index: - name: autocrop-tekst results: [] ---  # autocrop-tekst This model is a fine-tuned version of [nvidia/mit-b0](https://huggingface.co/nvidia/mit-b0) on the /mnt/disk1/autocrop-data/datasets/tekst dataset. It achieves the following results on the evaluation set: - Loss: 0.0197 - Mean Iou: 0.4970 - Mean Accuracy: 0.9939 - Overall Accuracy: 0.9939 - Accuracy Background: nan - Accuracy Crop: 0.9939 - Iou Background: 0.0 - Iou Crop: 0.9939 ## 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: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments - lr_scheduler_type: cosine - lr_scheduler_warmup_steps: 0.1 - num_epochs: 50.0 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Mean Iou | Mean Accuracy | Overall Accuracy | Accuracy Background | Accuracy Crop | Iou Background | Iou Crop | |:-------------:|:-----:|:-----:|:---------------:|:--------:|:-------------:|:----------------:|:-------------------:|:-------------:|:--------------:|:--------:| | 0.1117 | 1.0 | 624 | 0.0847 | 0.4905 | 0.9810 | 0.9810 | nan | 0.9810 | 0.0 | 0.9810 | | 0.0532 | 2.0 | 1248 | 0.0444 | 0.4940 | 0.9880 | 0.9880 | nan | 0.9880 | 0.0 | 0.9880 | | 0.0382 | 3.0 | 1872 | 0.0330 | 0.4962 | 0.9924 | 0.9924 | nan | 0.9924 | 0.0 | 0.9924 | | 0.0302 | 4.0 | 2496 | 0.0292 | 0.4954 | 0.9908 | 0.9908 | nan | 0.9908 | 0.0 | 0.9908 | | 0.0294 | 5.0 | 3120 | 0.0279 | 0.4946 | 0.9892 | 0.9892 | nan | 0.9892 | 0.0 | 0.9892 | | 0.0284 | 6.0 | 3744 | 0.0255 | 0.4962 | 0.9925 | 0.9925 | nan | 0.9925 | 0.0 | 0.9925 | | 0.0256 | 7.0 | 4368 | 0.0249 | 0.4953 | 0.9906 | 0.9906 | nan | 0.9906 | 0.0 | 0.9906 | | 0.0294...

NoteSep 1huggingface.co/NbAiLab/autocrop-tekst

Source context: 3 downloads · 0 likes · Pipeline image-segmentation · Library transformers · Repo NbAiLab/autocrop-tekst