--- license: apache-2.0 base_model: microsoft/beit-base-finetuned-ade-640-640 tags: - generated_from_trainer model-index: - name: BEiT_beit-base-finetuned-ade-640-640_Clean-Set3_RGB results: [] pipeline_tag: image-segmentation --- # BEiT_beit-base-finetuned-ade-640-640_Clean-Set3_RGB This model is a fine-tuned version of [microsoft/beit-base-finetuned-ade-640-640](https://huggingface.co/microsoft/beit-base-finetuned-ade-640-640) on an unknown dataset. It achieves the following results on the evaluation set: - TrainLoss: 0.0216 - Loss: 0.0336 - Mean Iou: 0.9671 - Mean Accuracy: 0.9806 - Overall Accuracy: 0.9926 - Accuracy Background: 0.9956 - Accuracy Melt: 0.9505 - Accuracy Substrate: 0.9957 - Iou Background: 0.9916 - Iou Melt: 0.9208 - Iou Substrate: 0.9888 ## 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: 2e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: cosine - lr_scheduler_warmup_steps: 200 - num_epochs: 50 ### Training results | Training Loss | Epoch | Step | Validation Loss | Mean Iou | Mean Accuracy | Overall Accuracy | Accuracy Background | Accuracy Melt | Accuracy Substrate | Iou Background | Iou Melt | Iou Substrate | |:-------------:|:-------:|:----:|:---------------:|:--------:|:-------------:|:----------------:|:-------------------:|:-------------:|:------------------:|:--------------:|:--------:|:-------------:| | 0.4042 | 0.9434 | 50 | 0.3272 | 0.8363 | 0.8672 | 0.9671 | 0.9931 | 0.6175 | 0.9911 | 0.9836 | 0.5790 | 0.9463 | | 0.1649 | 1.8868 | 100 | 0.0973 | 0.9371 | 0.9572 | 0.9867 | 0.9959 | 0.8833 | 0.9926 | 0.9881 | 0.8437 | 0.9795 | | 0.1439 | 2.8302 | 150 | 0.0724 | 0.9495 | 0.9800 | 0.9887 | 0.9946 | 0.9575 | 0.9879 | 0.9898 | 0.8770 | 0.9818 | | 0.1275 | 3.7736 | 200 | 0.0656 | 0.9443 | 0.9778 | 0.9877 | 0.9969 | 0.9515 | 0.9850 | 0.9903 | 0.8627 | 0.9799 | | 0.1522 | 4.7170 | 250 | 0.0585 | 0.9567 | 0.9737 | 0.9899 | 0.9971 | 0.9325 | 0.9915 | 0.9887 | 0.8976 | 0.9839 | | 0.1292 | 5.6604 | 300 | 0.0594 | 0.9502 | 0.9748 | 0.9877 | 0.9934 | 0.9418 | 0.9890 | 0.9857 | 0.8850 | 0.9801 | | 0.097 | 6.6038 | 350 | 0.0450 | 0.9634 | 0.9775 | 0.9912 | 0.9949 | 0.9432 | 0.9943 | 0.9883 | 0.9154 | 0.9866 | | 0.1125 | 7.5472 | 400 | 0.0451 | 0.9605 | 0.9757 | 0.9905 | 0.9953 | 0.9384 | 0.9934 | 0.9877 | 0.9080 | 0.9857 | | 0.102 | 8.4906 | 450 | 0.0518 | 0.9531 | 0.9798 | 0.9876 | 0.9921 | 0.9596 | 0.9876 | 0.9824 | 0.8960 | 0.9808 | | 0.0878 | 9.4340 | 500 | 0.0411 | 0.9639 | 0.9820 | 0.9911 | 0.9947 | 0.9592 | 0.9922 | 0.9885 | 0.9172 | 0.9859 | | 0.1198 | 10.3774 | 550 | 0.0679 | 0.9398 | 0.9655 | 0.9821 | 0.9873 | 0.9237 | 0.9855 | 0.9708 | 0.8768 | 0.9719 | | 0.055 | 11.3208 | 600 | 0.0521 | 0.9518 | 0.9791 | 0.9867 | 0.9846 | 0.9610 | 0.9917 | 0.9780 | 0.8966 | 0.9810 | | 0.086 | 12.2642 | 650 | 0.0402 | 0.9631 | 0.9791 | 0.9903 | 0.9920 | 0.9514 | 0.9940 | 0.9861 | 0.9185 | 0.9848 | | 0.058 | 13.2075 | 700 | 0.0455 | 0.9590 | 0.9768 | 0.9892 | 0.9908 | 0.9463 | 0.9934 | 0.9837 | 0.9096 | 0.9836 | | 0.0494 | 14.1509 | 750 | 0.0441 | 0.9588 | 0.9796 | 0.9895 | 0.9926 | 0.9547 | 0.9914 | 0.9842 | 0.9076 | 0.9846 | | 0.0599 | 15.0943 | 800 | 0.0401 | 0.9622 | 0.9787 | 0.9904 | 0.9925 | 0.9496 | 0.9939 | 0.9865 | 0.9149 | 0.9851 | | 0.0422 | 16.0377 | 850 | 0.0393 | 0.9619 | 0.9807 | 0.9906 | 0.9946 | 0.9556 | 0.9919 | 0.9880 | 0.9123 | 0.9853 | | 0.0454 | 16.9811 | 900 | 0.0429 | 0.9579 | 0.9742 | 0.9897 | 0.9918 | 0.9360 | 0.9948 | 0.9857 | 0.9033 | 0.9846 | | 0.0806 | 17.9245 | 950 | 0.0377 | 0.9640 | 0.9779 | 0.9915 | 0.9928 | 0.9445 | 0.9964 | 0.9892 | 0.9157 | 0.9869 | | 0.0677 | 18.8679 | 1000 | 0.0380 | 0.9602 | 0.9797 | 0.9910 | 0.9941 | 0.9513 | 0.9937 | 0.9882 | 0.9047 | 0.9877 | | 0.036 | 19.8113 | 1050 | 0.0388 | 0.9618 | 0.9799 | 0.9906 | 0.9942 | 0.9529 | 0.9925 | 0.9868 | 0.9127 | 0.9860 | | 0.0424 | 20.7547 | 1100 | 0.0375 | 0.9601 | 0.9753 | 0.9905 | 0.9934 | 0.9376 | 0.9949 | 0.9868 | 0.9071 | 0.9863 | | 0.0274 | 21.6981 | 1150 | 0.0322 | 0.9675 | 0.9795 | 0.9927 | 0.9955 | 0.9464 | 0.9965 | 0.9917 | 0.9218 | 0.9890 | | 0.0622 | 22.6415 | 1200 | 0.0360 | 0.9648 | 0.9798 | 0.9913 | 0.9932 | 0.9512 | 0.9949 | 0.9881 | 0.9197 | 0.9868 | | 0.0296 | 23.5849 | 1250 | 0.0334 | 0.9670 | 0.9823 | 0.9925 | 0.9953 | 0.9567 | 0.9950 | 0.9917 | 0.9207 | 0.9885 | | 0.0222 | 24.5283 | 1300 | 0.0326 | 0.9674 | 0.9823 | 0.9925 | 0.9948 | 0.9569 | 0.9953 | 0.9912 | 0.9222 | 0.9887 | | 0.0719 | 25.4717 | 1350 | 0.0328 | 0.9671 | 0.9832 | 0.9923 | 0.9945 | 0.9603 | 0.9947 | 0.9907 | 0.9223 | 0.9883 | | 0.0197 | 26.4151 | 1400 | 0.0311 | 0.9681 | 0.9817 | 0.9929 | 0.9962 | 0.9537 | 0.9954 | 0.9922 | 0.9230 | 0.9893 | | 0.0223 | 27.3585 | 1450 | 0.0324 | 0.9664 | 0.9811 | 0.9925 | 0.9956 | 0.9527 | 0.9950 | 0.9916 | 0.9191 | 0.9885 | | 0.024 | 28.3019 | 1500 | 0.0340 | 0.9657 | 0.9808 | 0.9920 | 0.9950 | 0.9528 | 0.9947 | 0.9902 | 0.9190 | 0.9880 | | 0.0242 | 29.2453 | 1550 | 0.0325 | 0.9672 | 0.9810 | 0.9926 | 0.9953 | 0.9522 | 0.9957 | 0.9915 | 0.9212 | 0.9888 | | 0.0371 | 30.1887 | 1600 | 0.0315 | 0.9681 | 0.9826 | 0.9928 | 0.9957 | 0.9569 | 0.9952 | 0.9920 | 0.9232 | 0.9891 | | 0.0235 | 31.1321 | 1650 | 0.0370 | 0.9632 | 0.9799 | 0.9911 | 0.9937 | 0.9520 | 0.9941 | 0.9880 | 0.9150 | 0.9868 | | 0.0266 | 32.0755 | 1700 | 0.0335 | 0.9664 | 0.9811 | 0.9925 | 0.9951 | 0.9527 | 0.9954 | 0.9913 | 0.9193 | 0.9887 | | 0.0216 | 33.0189 | 1750 | 0.0344 | 0.9656 | 0.9800 | 0.9921 | 0.9946 | 0.9497 | 0.9956 | 0.9904 | 0.9182 | 0.9883 | | 0.0382 | 33.9623 | 1800 | 0.0319 | 0.9680 | 0.9819 | 0.9929 | 0.9954 | 0.9544 | 0.9959 | 0.9922 | 0.9224 | 0.9893 | | 0.0161 | 34.9057 | 1850 | 0.0336 | 0.9672 | 0.9799 | 0.9927 | 0.9955 | 0.9479 | 0.9963 | 0.9920 | 0.9206 | 0.9890 | | 0.0216 | 35.8491 | 1900 | 0.0336 | 0.9671 | 0.9806 | 0.9926 | 0.9956 | 0.9505 | 0.9957 | 0.9916 | 0.9208 | 0.9888 | ### Framework versions - Transformers 4.41.2 - Pytorch 2.0.1+cu117 - Datasets 2.19.2 - Tokenizers 0.19.1