segformer-b1-finetuned-segments-graffiti

This model is a fine-tuned version of nvidia/mit-b1 on the Adriatogi/graffiti dataset. It achieves the following results on the evaluation set:

• Loss: 0.2171
• Mean Iou: 0.8381
• Mean Accuracy: 0.9102
• Overall Accuracy: 0.9168
• Accuracy Not Graf: 0.9379
• Accuracy Graf: 0.8826
• Iou Not Graf: 0.8748
• Iou Graf: 0.8015

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Not Graf	Accuracy Graf	Iou Not Graf	Iou Graf
0.4076	0.42	20	0.5389	0.6053	0.7982	0.7541	0.6139	0.9825	0.6073	0.6033
0.3386	0.83	40	0.2883	0.7962	0.8984	0.8898	0.8625	0.9343	0.8290	0.7634
0.1964	1.25	60	0.2514	0.8061	0.9009	0.8964	0.8819	0.9200	0.8406	0.7716
0.1723	1.67	80	0.2259	0.8269	0.9058	0.9100	0.9235	0.8880	0.8641	0.7898
0.1981	2.08	100	0.2338	0.8119	0.9040	0.8999	0.8869	0.9210	0.8459	0.7778
0.2827	2.5	120	0.2106	0.8251	0.9080	0.9084	0.9095	0.9066	0.8601	0.7902
0.1864	2.92	140	0.2241	0.8232	0.8956	0.9097	0.9546	0.8365	0.8675	0.7790
0.1362	3.33	160	0.2185	0.8257	0.8978	0.9109	0.9525	0.8431	0.8688	0.7826
0.1264	3.75	180	0.2155	0.8237	0.9054	0.9079	0.9156	0.8952	0.8602	0.7871
0.1688	4.17	200	0.2241	0.8206	0.8985	0.9072	0.9346	0.8625	0.8618	0.7795
0.1198	4.58	220	0.2080	0.8331	0.9087	0.9137	0.9296	0.8877	0.8697	0.7965
0.111	5.0	240	0.2033	0.8369	0.9133	0.9154	0.9221	0.9044	0.8710	0.8027
0.2003	5.42	260	0.2214	0.8262	0.9118	0.9084	0.8976	0.9261	0.8586	0.7938
0.1369	5.83	280	0.2044	0.8396	0.9147	0.9170	0.9245	0.9048	0.8734	0.8058
0.1901	6.25	300	0.1968	0.8411	0.9119	0.9185	0.9393	0.8846	0.8771	0.8050
0.1887	6.67	320	0.2098	0.8367	0.9100	0.9159	0.9344	0.8857	0.8731	0.8002
0.0738	7.08	340	0.2205	0.8357	0.9127	0.9147	0.9211	0.9043	0.8699	0.8014
0.1166	7.5	360	0.2274	0.8317	0.9046	0.9135	0.9420	0.8672	0.8709	0.7924
0.1247	7.92	380	0.2225	0.8310	0.9051	0.9130	0.9381	0.8722	0.8698	0.7923
0.1212	8.33	400	0.2230	0.8345	0.9108	0.9143	0.9254	0.8961	0.8699	0.7991
0.0979	8.75	420	0.2226	0.8352	0.9076	0.9153	0.9400	0.8752	0.8730	0.7973
0.0984	9.17	440	0.2189	0.8354	0.9106	0.9149	0.9287	0.8925	0.8712	0.7997
0.1151	9.58	460	0.2185	0.8382	0.9098	0.9170	0.9396	0.8800	0.8751	0.8013
0.0989	10.0	480	0.2171	0.8381	0.9102	0.9168	0.9379	0.8826	0.8748	0.8015

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2