segformer-finetuned-biofilm_MRCNNv1_concat

This model is a fine-tuned version of nvidia/mit-b0 on the heroza/biofilm_MRCNNv1_concat dataset. It achieves the following results on the evaluation set:

• Loss: 0.0530
• Mean Iou: 0.8677
• Mean Accuracy: 0.9780
• Overall Accuracy: 0.9815
• Accuracy Background: 0.9820
• Accuracy Biofilm: 0.9740
• Iou Background: 0.9804
• Iou Biofilm: 0.7549

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
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Biofilm	Iou Background	Iou Biofilm
0.0759	1.0	433	0.0649	0.8684	0.9740	0.9818	0.9829	0.9651	0.9807	0.7561
0.0435	2.0	866	0.0391	0.8875	0.9616	0.9855	0.9887	0.9345	0.9847	0.7903
0.0407	3.0	1299	0.0353	0.8951	0.9696	0.9865	0.9888	0.9504	0.9857	0.8046
0.0372	4.0	1732	0.0489	0.8765	0.9810	0.9830	0.9833	0.9788	0.9820	0.7711
0.0378	5.0	2165	0.0311	0.9020	0.9574	0.9879	0.9919	0.9229	0.9872	0.8168
0.0325	6.0	2598	0.0510	0.8663	0.9745	0.9814	0.9823	0.9666	0.9803	0.7524
0.0306	7.0	3031	0.0428	0.8873	0.9760	0.9850	0.9862	0.9657	0.9842	0.7904
0.0318	8.0	3464	0.0399	0.8837	0.9739	0.9845	0.9859	0.9618	0.9836	0.7839
0.0302	9.0	3897	0.0436	0.8795	0.9689	0.9840	0.9859	0.9520	0.9830	0.7760
0.0236	10.0	4330	0.0391	0.8856	0.9713	0.9849	0.9867	0.9560	0.9840	0.7871
0.0247	11.0	4763	0.0451	0.8705	0.9731	0.9822	0.9834	0.9628	0.9812	0.7598
0.0213	12.0	5196	0.0487	0.8656	0.9735	0.9813	0.9824	0.9647	0.9802	0.7510
0.0256	13.0	5629	0.0444	0.8799	0.9668	0.9841	0.9864	0.9473	0.9832	0.7766
0.0218	14.0	6062	0.0492	0.8679	0.9773	0.9816	0.9822	0.9725	0.9805	0.7553
0.0216	15.0	6495	0.0502	0.8717	0.9748	0.9824	0.9834	0.9663	0.9813	0.7621
0.0206	16.0	6928	0.0565	0.8623	0.9766	0.9806	0.9811	0.9721	0.9794	0.7453
0.0223	17.0	7361	0.0509	0.8666	0.9730	0.9815	0.9826	0.9635	0.9804	0.7527
0.0226	18.0	7794	0.0464	0.8794	0.9792	0.9836	0.9842	0.9743	0.9826	0.7762
0.0243	19.0	8227	0.0546	0.8649	0.9824	0.9809	0.9806	0.9843	0.9797	0.7501
0.02	20.0	8660	0.0567	0.8648	0.9766	0.9810	0.9816	0.9716	0.9799	0.7496
0.0196	21.0	9093	0.0559	0.8648	0.9784	0.9810	0.9813	0.9755	0.9798	0.7497
0.0206	22.0	9526	0.0552	0.8652	0.9779	0.9811	0.9815	0.9742	0.9799	0.7504
0.0189	23.0	9959	0.0544	0.8661	0.9785	0.9812	0.9816	0.9753	0.9801	0.7521
0.0208	23.09	10000	0.0530	0.8677	0.9780	0.9815	0.9820	0.9740	0.9804	0.7549

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.0.0+cu117
• Datasets 2.14.4
• Tokenizers 0.15.1