segformer-finetuned-biofilm2_train

This model is a fine-tuned version of nvidia/mit-b0 on the heroza/biofilm2_train dataset. It achieves the following results on the evaluation set:

• Loss: 0.0761
• Mean Iou: 0.8665
• Mean Accuracy: 0.9765
• Overall Accuracy: 0.9745
• Accuracy Background: 0.9741
• Accuracy Biofilm: 0.9789
• Iou Background: 0.9722
• Iou Biofilm: 0.7608

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
0.1611	1.0	298	0.1220	0.8393	0.9547	0.9687	0.9714	0.9379	0.9660	0.7126
0.07	2.0	596	0.0682	0.8795	0.9359	0.9795	0.9881	0.8837	0.9779	0.7811
0.0542	3.0	894	0.0564	0.8862	0.9735	0.9793	0.9805	0.9666	0.9775	0.7948
0.0508	4.0	1192	0.0517	0.8888	0.9728	0.9799	0.9814	0.9643	0.9782	0.7993
0.0491	5.0	1490	0.0479	0.8999	0.9727	0.9824	0.9843	0.9611	0.9809	0.8190
0.0496	6.0	1788	0.0665	0.8733	0.9728	0.9764	0.9770	0.9686	0.9743	0.7724
0.047	7.0	2086	0.0475	0.8936	0.9744	0.9810	0.9823	0.9664	0.9793	0.8079
0.0403	8.0	2384	0.0513	0.8897	0.9699	0.9803	0.9823	0.9575	0.9786	0.8008
0.0336	9.0	2682	0.0597	0.8736	0.9790	0.9761	0.9756	0.9824	0.9740	0.7732
0.036	10.0	2980	0.0602	0.8789	0.9781	0.9774	0.9773	0.9789	0.9755	0.7824
0.0335	11.0	3278	0.0519	0.8849	0.9670	0.9793	0.9818	0.9522	0.9775	0.7923
0.0364	12.0	3576	0.0684	0.8718	0.9810	0.9756	0.9745	0.9874	0.9734	0.7702
0.0423	13.0	3874	0.0637	0.8767	0.9742	0.9771	0.9777	0.9707	0.9751	0.7783
0.0354	14.0	4172	0.0618	0.8773	0.9692	0.9775	0.9791	0.9593	0.9755	0.7790
0.0335	15.0	4470	0.0547	0.8788	0.9686	0.9778	0.9797	0.9574	0.9759	0.7816
0.0318	16.0	4768	0.0567	0.8841	0.9744	0.9788	0.9797	0.9691	0.9770	0.7913
0.0296	17.0	5066	0.0653	0.8678	0.9741	0.9749	0.9751	0.9732	0.9727	0.7628
0.0291	18.0	5364	0.0591	0.8757	0.9718	0.9770	0.9780	0.9657	0.9750	0.7765
0.0311	19.0	5662	0.0716	0.8682	0.9753	0.9750	0.9749	0.9756	0.9728	0.7637
0.0322	20.0	5960	0.0837	0.8506	0.9773	0.9703	0.9690	0.9857	0.9677	0.7335
0.0317	21.0	6258	0.0728	0.8673	0.9749	0.9748	0.9747	0.9751	0.9726	0.7621
0.0318	22.0	6556	0.0571	0.8796	0.9764	0.9777	0.9779	0.9748	0.9757	0.7835
0.0288	23.0	6854	0.0734	0.8689	0.9798	0.9749	0.9739	0.9858	0.9727	0.7651
0.0271	24.0	7152	0.0763	0.8615	0.9757	0.9733	0.9728	0.9785	0.9709	0.7521
0.0236	25.0	7450	0.0615	0.8789	0.9761	0.9775	0.9778	0.9744	0.9756	0.7823
0.025	26.0	7748	0.0694	0.8684	0.9768	0.9750	0.9746	0.9790	0.9727	0.7640
0.0269	27.0	8046	0.0672	0.8700	0.9688	0.9757	0.9771	0.9605	0.9736	0.7664
0.0286	28.0	8344	0.0717	0.8695	0.9761	0.9753	0.9751	0.9771	0.9731	0.7659
0.0255	29.0	8642	0.0680	0.8696	0.9757	0.9753	0.9752	0.9761	0.9731	0.7661
0.0255	30.0	8940	0.0701	0.8691	0.9756	0.9752	0.9751	0.9762	0.9730	0.7651
0.0223	31.0	9238	0.0715	0.8687	0.9746	0.9751	0.9752	0.9740	0.9730	0.7644
0.0226	32.0	9536	0.0757	0.8667	0.9770	0.9745	0.9740	0.9799	0.9723	0.7612
0.022	33.0	9834	0.0773	0.8661	0.9766	0.9744	0.9739	0.9793	0.9721	0.7601
0.0217	33.56	10000	0.0761	0.8665	0.9765	0.9745	0.9741	0.9789	0.9722	0.7608

Framework versions

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