segformer-finetuned-biofilm_MRCNNv1_train80_val20
This model is a fine-tuned version of nvidia/mit-b0 on the heroza/biofilm_MRCNNv1_train80_val20 dataset. It achieves the following results on the evaluation set:
- Loss: 0.0785
- Mean Iou: 0.8728
- Mean Accuracy: 0.9695
- Overall Accuracy: 0.9773
- Accuracy Background: 0.9788
- Accuracy Biofilm: 0.9603
- Iou Background: 0.9755
- Iou Biofilm: 0.7702
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: 1337
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: polynomial
- num_epochs: 50.0
Training results
Training Loss | Epoch | Step | Validation Loss | Mean Iou | Mean Accuracy | Overall Accuracy | Accuracy Background | Accuracy Biofilm | Iou Background | Iou Biofilm |
---|---|---|---|---|---|---|---|---|---|---|
0.1517 | 1.0 | 280 | 0.1319 | 0.8108 | 0.8974 | 0.9664 | 0.9794 | 0.8154 | 0.9641 | 0.6575 |
0.0728 | 2.0 | 560 | 0.0636 | 0.8857 | 0.9582 | 0.9807 | 0.9849 | 0.9314 | 0.9791 | 0.7923 |
0.0555 | 3.0 | 840 | 0.0526 | 0.8863 | 0.9417 | 0.9815 | 0.9889 | 0.8944 | 0.9800 | 0.7925 |
0.045 | 4.0 | 1120 | 0.0522 | 0.8763 | 0.9225 | 0.9802 | 0.9911 | 0.8539 | 0.9788 | 0.7738 |
0.048 | 5.0 | 1400 | 0.0415 | 0.9085 | 0.9742 | 0.9848 | 0.9868 | 0.9616 | 0.9835 | 0.8335 |
0.0425 | 6.0 | 1680 | 0.0468 | 0.9041 | 0.9790 | 0.9837 | 0.9846 | 0.9733 | 0.9824 | 0.8258 |
0.0438 | 7.0 | 1960 | 0.0461 | 0.9001 | 0.9717 | 0.9832 | 0.9853 | 0.9581 | 0.9818 | 0.8185 |
0.0406 | 8.0 | 2240 | 0.0488 | 0.8936 | 0.9701 | 0.9819 | 0.9841 | 0.9561 | 0.9804 | 0.8068 |
0.0354 | 9.0 | 2520 | 0.0434 | 0.9004 | 0.9648 | 0.9835 | 0.9870 | 0.9426 | 0.9821 | 0.8187 |
0.0411 | 10.0 | 2800 | 0.0467 | 0.9021 | 0.9680 | 0.9837 | 0.9867 | 0.9494 | 0.9824 | 0.8219 |
0.0385 | 11.0 | 3080 | 0.0846 | 0.8493 | 0.9697 | 0.9716 | 0.9720 | 0.9674 | 0.9692 | 0.7294 |
0.0327 | 12.0 | 3360 | 0.0574 | 0.8709 | 0.9643 | 0.9771 | 0.9795 | 0.9491 | 0.9753 | 0.7666 |
0.0333 | 13.0 | 3640 | 0.0654 | 0.8649 | 0.9645 | 0.9757 | 0.9778 | 0.9513 | 0.9737 | 0.7561 |
0.0356 | 14.0 | 3920 | 0.0823 | 0.8472 | 0.9703 | 0.9710 | 0.9712 | 0.9694 | 0.9686 | 0.7259 |
0.0277 | 15.0 | 4200 | 0.0657 | 0.8634 | 0.9761 | 0.9748 | 0.9745 | 0.9777 | 0.9727 | 0.7542 |
0.0328 | 16.0 | 4480 | 0.0575 | 0.8785 | 0.9668 | 0.9787 | 0.9810 | 0.9526 | 0.9770 | 0.7800 |
0.0362 | 17.0 | 4760 | 0.0595 | 0.8750 | 0.9696 | 0.9778 | 0.9794 | 0.9599 | 0.9760 | 0.7741 |
0.0301 | 18.0 | 5040 | 0.0610 | 0.8701 | 0.9755 | 0.9764 | 0.9766 | 0.9744 | 0.9744 | 0.7659 |
0.0284 | 19.0 | 5320 | 0.0562 | 0.8874 | 0.9748 | 0.9804 | 0.9814 | 0.9682 | 0.9787 | 0.7961 |
0.0345 | 20.0 | 5600 | 0.0612 | 0.8704 | 0.9696 | 0.9767 | 0.9781 | 0.9611 | 0.9748 | 0.7659 |
0.0292 | 21.0 | 5880 | 0.0689 | 0.8639 | 0.9728 | 0.9751 | 0.9755 | 0.9701 | 0.9730 | 0.7549 |
0.0287 | 22.0 | 6160 | 0.0559 | 0.8821 | 0.9657 | 0.9796 | 0.9822 | 0.9493 | 0.9779 | 0.7862 |
0.0307 | 23.0 | 6440 | 0.0683 | 0.8637 | 0.9716 | 0.9751 | 0.9757 | 0.9674 | 0.9730 | 0.7545 |
0.0303 | 24.0 | 6720 | 0.0692 | 0.8702 | 0.9627 | 0.9770 | 0.9797 | 0.9457 | 0.9752 | 0.7653 |
0.0257 | 25.0 | 7000 | 0.0594 | 0.8783 | 0.9709 | 0.9785 | 0.9799 | 0.9619 | 0.9767 | 0.7798 |
0.0341 | 26.0 | 7280 | 0.0762 | 0.8619 | 0.9746 | 0.9745 | 0.9745 | 0.9747 | 0.9723 | 0.7515 |
0.025 | 27.0 | 7560 | 0.0675 | 0.8696 | 0.9751 | 0.9763 | 0.9765 | 0.9736 | 0.9743 | 0.7649 |
0.0281 | 28.0 | 7840 | 0.0661 | 0.8641 | 0.9694 | 0.9753 | 0.9764 | 0.9625 | 0.9732 | 0.7550 |
0.0285 | 29.0 | 8120 | 0.0796 | 0.8592 | 0.9737 | 0.9739 | 0.9739 | 0.9736 | 0.9717 | 0.7467 |
0.0263 | 30.0 | 8400 | 0.0760 | 0.8627 | 0.9712 | 0.9749 | 0.9755 | 0.9668 | 0.9728 | 0.7527 |
0.0252 | 31.0 | 8680 | 0.0615 | 0.8800 | 0.9642 | 0.9792 | 0.9820 | 0.9464 | 0.9775 | 0.7825 |
0.0245 | 32.0 | 8960 | 0.0647 | 0.8665 | 0.9642 | 0.9761 | 0.9783 | 0.9501 | 0.9742 | 0.7589 |
0.0241 | 33.0 | 9240 | 0.0638 | 0.8749 | 0.9668 | 0.9779 | 0.9800 | 0.9535 | 0.9761 | 0.7736 |
0.0249 | 34.0 | 9520 | 0.0803 | 0.8610 | 0.9709 | 0.9744 | 0.9751 | 0.9667 | 0.9723 | 0.7497 |
0.0213 | 35.0 | 9800 | 0.0754 | 0.8687 | 0.9633 | 0.9767 | 0.9792 | 0.9474 | 0.9748 | 0.7627 |
0.0233 | 36.0 | 10080 | 0.0675 | 0.8743 | 0.9612 | 0.9780 | 0.9812 | 0.9411 | 0.9763 | 0.7723 |
0.0214 | 37.0 | 10360 | 0.0695 | 0.8758 | 0.9714 | 0.9779 | 0.9791 | 0.9637 | 0.9761 | 0.7755 |
0.0231 | 38.0 | 10640 | 0.0704 | 0.8695 | 0.9621 | 0.9769 | 0.9797 | 0.9444 | 0.9750 | 0.7640 |
0.0231 | 39.0 | 10920 | 0.0780 | 0.8636 | 0.9646 | 0.9754 | 0.9774 | 0.9518 | 0.9734 | 0.7539 |
0.0221 | 40.0 | 11200 | 0.0726 | 0.8709 | 0.9666 | 0.9770 | 0.9790 | 0.9542 | 0.9751 | 0.7666 |
0.0227 | 41.0 | 11480 | 0.0829 | 0.8618 | 0.9627 | 0.9751 | 0.9774 | 0.9480 | 0.9730 | 0.7505 |
0.0241 | 42.0 | 11760 | 0.0701 | 0.8763 | 0.9679 | 0.9782 | 0.9801 | 0.9557 | 0.9764 | 0.7762 |
0.0206 | 43.0 | 12040 | 0.0782 | 0.8666 | 0.9670 | 0.9760 | 0.9777 | 0.9563 | 0.9740 | 0.7593 |
0.023 | 44.0 | 12320 | 0.0809 | 0.8656 | 0.9654 | 0.9758 | 0.9778 | 0.9530 | 0.9739 | 0.7573 |
0.0223 | 45.0 | 12600 | 0.0805 | 0.8688 | 0.9660 | 0.9765 | 0.9785 | 0.9535 | 0.9746 | 0.7630 |
0.0224 | 46.0 | 12880 | 0.0748 | 0.8719 | 0.9682 | 0.9772 | 0.9789 | 0.9576 | 0.9753 | 0.7685 |
0.0233 | 47.0 | 13160 | 0.0796 | 0.8697 | 0.9669 | 0.9767 | 0.9786 | 0.9552 | 0.9748 | 0.7645 |
0.019 | 48.0 | 13440 | 0.0772 | 0.8729 | 0.9681 | 0.9774 | 0.9792 | 0.9569 | 0.9756 | 0.7703 |
0.0215 | 49.0 | 13720 | 0.0783 | 0.8720 | 0.9682 | 0.9772 | 0.9789 | 0.9575 | 0.9753 | 0.7686 |
0.0186 | 50.0 | 14000 | 0.0785 | 0.8728 | 0.9695 | 0.9773 | 0.9788 | 0.9603 | 0.9755 | 0.7702 |
Framework versions
- Transformers 4.38.0.dev0
- Pytorch 2.0.0+cu117
- Datasets 2.14.4
- Tokenizers 0.15.1
- Downloads last month
- 4
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for heroza/segformer-finetuned-biofilm_MRCNNv1_train80_val20
Base model
nvidia/mit-b0