segformer-b0-finetuned-raw_img_ready2train_patches

This model is a fine-tuned version of nvidia/mit-b0 on the raw_img_ready2train_patches dataset. It achieves the following results on the evaluation set:

• Loss: 0.6829
• Mean Iou: 0.4110
• Mean Accuracy: 0.7629
• Overall Accuracy: 0.7631
• Accuracy Unlabeled: nan
• Accuracy Eczema: 0.7673
• Accuracy Background: 0.7585
• Iou Unlabeled: 0.0
• Iou Eczema: 0.6284
• Iou Background: 0.6047

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: 5e-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: linear
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Eczema	Accuracy Background	Iou Unlabeled	Iou Eczema	Iou Background
1.0753	0.0312	5	1.0925	0.2358	0.4682	0.4698	nan	0.5042	0.4322	0.0	0.3705	0.3367
0.9863	0.0625	10	1.0697	0.2994	0.6182	0.6306	nan	0.8979	0.3385	0.0	0.5784	0.3198
1.0056	0.0938	15	1.0377	0.3303	0.6678	0.6792	nan	0.9236	0.4121	0.0	0.6064	0.3844
1.0133	0.125	20	1.0006	0.3478	0.6869	0.6950	nan	0.8710	0.5027	0.0	0.6008	0.4425
0.9748	0.1562	25	0.9689	0.3543	0.6947	0.7022	nan	0.8647	0.5246	0.0	0.6043	0.4586
0.9367	0.1875	30	0.9417	0.3566	0.6950	0.6965	nan	0.7290	0.6610	0.0	0.5583	0.5114
0.8363	0.2188	35	0.9118	0.3557	0.6940	0.6959	nan	0.7366	0.6514	0.0	0.5600	0.5069
1.1431	0.25	40	0.8830	0.3575	0.6963	0.6989	nan	0.7556	0.6370	0.0	0.5686	0.5039
0.7312	0.2812	45	0.8592	0.3680	0.7098	0.7133	nan	0.7888	0.6307	0.0	0.5907	0.5133
0.8135	0.3125	50	0.8268	0.3559	0.6994	0.7083	nan	0.8992	0.4997	0.0	0.6173	0.4505
0.7528	0.3438	55	0.8110	0.3525	0.6960	0.7053	nan	0.9055	0.4866	0.0	0.6162	0.4412
0.8405	0.375	60	0.7967	0.3518	0.6950	0.7041	nan	0.9008	0.4893	0.0	0.6140	0.4415
0.7865	0.4062	65	0.7791	0.3561	0.6992	0.7075	nan	0.8869	0.5116	0.0	0.6130	0.4553
0.8309	0.4375	70	0.7650	0.3652	0.7083	0.7147	nan	0.8512	0.5655	0.0	0.6090	0.4864
0.6775	0.4688	75	0.7615	0.3613	0.7044	0.7115	nan	0.8651	0.5437	0.0	0.6102	0.4738
0.7033	0.5	80	0.7498	0.3737	0.7179	0.7227	nan	0.8260	0.6099	0.0	0.6087	0.5125
0.8377	0.5312	85	0.7443	0.3790	0.7243	0.7290	nan	0.8303	0.6184	0.0	0.6154	0.5217
0.825	0.5625	90	0.7547	0.3676	0.7125	0.7201	nan	0.8840	0.5411	0.0	0.6225	0.4802
0.7408	0.5938	95	0.7415	0.3767	0.7228	0.7295	nan	0.8747	0.5708	0.0	0.6281	0.5021
0.8087	0.625	100	0.7201	0.3926	0.7404	0.7445	nan	0.8318	0.6491	0.0	0.6296	0.5483
0.7146	0.6562	105	0.7096	0.4002	0.7493	0.7520	nan	0.8109	0.6877	0.0	0.6307	0.5699
0.6875	0.6875	110	0.7047	0.4010	0.7502	0.7541	nan	0.8398	0.6606	0.0	0.6407	0.5621
0.6382	0.7188	115	0.7031	0.3982	0.7471	0.7519	nan	0.8543	0.6400	0.0	0.6426	0.5521
0.6551	0.75	120	0.6953	0.4018	0.7512	0.7553	nan	0.8450	0.6573	0.0	0.6433	0.5621
0.7074	0.7812	125	0.6912	0.4054	0.7553	0.7583	nan	0.8236	0.6871	0.0	0.6402	0.5760
0.768	0.8125	130	0.6866	0.4048	0.7546	0.7579	nan	0.8278	0.6814	0.0	0.6410	0.5736
0.7543	0.8438	135	0.6851	0.4031	0.7526	0.7564	nan	0.8374	0.6679	0.0	0.6422	0.5671
0.7107	0.875	140	0.6803	0.6122	0.7586	0.7608	nan	0.8071	0.7101	nan	0.6379	0.5865
0.7054	0.9062	145	0.6799	0.4098	0.7608	0.7622	nan	0.7924	0.7292	0.0	0.6350	0.5943
1.1302	0.9375	150	0.6801	0.4103	0.7616	0.7626	nan	0.7840	0.7393	0.0	0.6330	0.5981
0.6037	0.9688	155	0.6827	0.4111	0.7628	0.7632	nan	0.7721	0.7534	0.0	0.6300	0.6032
0.8577	1.0	160	0.6829	0.4110	0.7629	0.7631	nan	0.7673	0.7585	0.0	0.6284	0.6047

Framework versions

• Transformers 4.40.1
• Pytorch 2.3.0
• Datasets 2.19.0
• Tokenizers 0.19.1