SegFormer_b2_2

This model is a fine-tuned version of nvidia/segformer-b2-finetuned-cityscapes-1024-1024 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5687
• Mean Iou: 0.7141
• Mean Accuracy: 0.8338
• Overall Accuracy: 0.9518
• Accuracy Road: 0.9861
• Accuracy Sidewalk: 0.9403
• Accuracy Building: 0.9549
• Accuracy Wall: 0.7046
• Accuracy Fence: 0.7106
• Accuracy Pole: 0.6880
• Accuracy Traffic light: 0.8719
• Accuracy Traffic sign: 0.8349
• Accuracy Vegetation: 0.9442
• Accuracy Terrain: 0.6876
• Accuracy Sky: 0.9817
• Accuracy Person: 0.8778
• Accuracy Rider: 0.5796
• Accuracy Car: 0.9746
• Accuracy Truck: 0.7663
• Accuracy Bus: 0.9041
• Accuracy Train: 0.7933
• Accuracy Motorcycle: 0.7614
• Accuracy Bicycle: 0.8798
• Iou Road: 0.9809
• Iou Sidewalk: 0.8418
• Iou Building: 0.9125
• Iou Wall: 0.5459
• Iou Fence: 0.5277
• Iou Pole: 0.5466
• Iou Traffic light: 0.6398
• Iou Traffic sign: 0.7499
• Iou Vegetation: 0.9115
• Iou Terrain: 0.5282
• Iou Sky: 0.9396
• Iou Person: 0.7568
• Iou Rider: 0.4521
• Iou Car: 0.9319
• Iou Truck: 0.6160
• Iou Bus: 0.7445
• Iou Train: 0.6995
• Iou Motorcycle: 0.5142
• Iou Bicycle: 0.7285

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: 0.0002
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 130
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Road	Accuracy Sidewalk	Accuracy Building	Accuracy Wall	Accuracy Fence	Accuracy Pole	Accuracy Traffic light	Accuracy Traffic sign	Accuracy Vegetation	Accuracy Terrain	Accuracy Sky	Accuracy Person	Accuracy Rider	Accuracy Car	Accuracy Truck	Accuracy Bus	Accuracy Train	Accuracy Motorcycle	Accuracy Bicycle	Iou Road	Iou Sidewalk	Iou Building	Iou Wall	Iou Fence	Iou Pole	Iou Traffic light	Iou Traffic sign	Iou Vegetation	Iou Terrain	Iou Sky	Iou Person	Iou Rider	Iou Car	Iou Truck	Iou Bus	Iou Train	Iou Motorcycle	Iou Bicycle
0.6593	0.2688	100	0.5767	0.7490	0.8653	0.9557	0.9879	0.9291	0.9527	0.6811	0.6907	0.7639	0.8966	0.8935	0.9505	0.7808	0.9856	0.9001	0.7336	0.9768	0.7932	0.9450	0.8737	0.8252	0.8802	0.9820	0.8569	0.9152	0.6022	0.5665	0.5525	0.6085	0.7337	0.9183	0.6399	0.9381	0.7709	0.5525	0.9413	0.7157	0.8153	0.8165	0.5705	0.7338
0.6791	0.5376	200	0.5698	0.7492	0.8600	0.9561	0.9880	0.9336	0.9583	0.6325	0.7103	0.7513	0.8553	0.8639	0.9489	0.7918	0.9852	0.8956	0.7517	0.9729	0.7818	0.9393	0.8832	0.8112	0.8840	0.9821	0.8560	0.9163	0.5714	0.5725	0.5599	0.6532	0.7479	0.9181	0.6287	0.9407	0.7748	0.5506	0.9428	0.6955	0.8014	0.8020	0.5780	0.7421
0.684	0.8065	300	0.5660	0.7540	0.8645	0.9570	0.9883	0.9310	0.9544	0.7076	0.7504	0.7494	0.8755	0.8758	0.9524	0.8069	0.9867	0.8878	0.7234	0.9740	0.7900	0.9457	0.8782	0.7426	0.9050	0.9827	0.8597	0.9187	0.6108	0.5737	0.5617	0.6427	0.7589	0.9198	0.6545	0.9430	0.7718	0.5517	0.9434	0.7264	0.7988	0.7947	0.5935	0.7189
0.658	1.0753	400	0.5648	0.7555	0.8713	0.9566	0.9876	0.9253	0.9519	0.7672	0.6932	0.7528	0.8699	0.8854	0.9538	0.7916	0.9818	0.9127	0.7634	0.9750	0.8688	0.9482	0.8984	0.7359	0.8924	0.9820	0.8559	0.9184	0.6372	0.5723	0.5547	0.6505	0.7551	0.9188	0.6353	0.9460	0.7588	0.5468	0.9409	0.7869	0.8348	0.7207	0.6050	0.7344
0.5832	1.3441	500	0.5662	0.7441	0.8714	0.9555	0.9888	0.9327	0.9422	0.7124	0.7149	0.7371	0.8847	0.8679	0.9616	0.7678	0.9862	0.8949	0.7883	0.9697	0.8754	0.9480	0.9008	0.7969	0.8863	0.9828	0.8533	0.9149	0.6197	0.5738	0.5542	0.6272	0.7509	0.9178	0.6262	0.9390	0.7630	0.5299	0.9410	0.7117	0.8474	0.7103	0.5715	0.7036
0.5949	1.6129	600	0.5658	0.7512	0.8710	0.9555	0.9878	0.9256	0.9433	0.7911	0.7468	0.7528	0.9006	0.8701	0.9563	0.7578	0.9817	0.9151	0.7248	0.9787	0.9052	0.9474	0.8130	0.7913	0.8603	0.9815	0.8545	0.9147	0.6437	0.5660	0.5541	0.6132	0.7441	0.9176	0.6410	0.9440	0.7493	0.5356	0.9394	0.7897	0.8492	0.7631	0.5436	0.7285
0.5894	1.8817	700	0.5636	0.7371	0.8666	0.9543	0.9869	0.9263	0.9460	0.7916	0.6749	0.7552	0.8802	0.8670	0.9504	0.8268	0.9835	0.9047	0.7147	0.9759	0.8055	0.9644	0.7929	0.8427	0.8767	0.9815	0.8548	0.9136	0.6382	0.5519	0.5478	0.6420	0.7505	0.9150	0.6215	0.9411	0.7632	0.5342	0.9389	0.7291	0.7657	0.7170	0.4891	0.7087
0.5715	2.1505	800	0.5679	0.7431	0.8700	0.9549	0.9862	0.9338	0.9460	0.7668	0.7404	0.7456	0.8754	0.8627	0.9608	0.7677	0.9812	0.8570	0.8195	0.9624	0.8757	0.9346	0.9068	0.7152	0.8921	0.9817	0.8523	0.9154	0.6357	0.5846	0.5438	0.6327	0.7417	0.9165	0.6558	0.9434	0.7518	0.4961	0.9354	0.6757	0.8227	0.7332	0.5868	0.7143
0.6365	2.4194	900	0.5647	0.7432	0.8610	0.9548	0.9870	0.9341	0.9505	0.6885	0.6909	0.7165	0.8781	0.8895	0.9547	0.7721	0.9881	0.8999	0.7258	0.9706	0.8670	0.9193	0.9065	0.7449	0.8744	0.9820	0.8544	0.9128	0.5677	0.5229	0.5553	0.6414	0.7557	0.9180	0.6489	0.9289	0.7704	0.5566	0.9367	0.7645	0.8216	0.7082	0.5442	0.7298
0.6795	2.6882	1000	0.5673	0.7301	0.8648	0.9525	0.9838	0.9305	0.9489	0.6736	0.7227	0.7042	0.9196	0.8746	0.9526	0.7629	0.9884	0.8850	0.7949	0.9674	0.8915	0.9001	0.8440	0.7920	0.8939	0.9788	0.8417	0.9104	0.5765	0.5507	0.5537	0.5471	0.7434	0.9160	0.6141	0.9285	0.7561	0.5163	0.9338	0.7324	0.8068	0.7732	0.4745	0.7187
0.6517	2.9570	1100	0.5647	0.7173	0.8507	0.9512	0.9836	0.9402	0.9490	0.7080	0.6173	0.7459	0.8519	0.8803	0.9454	0.7947	0.9821	0.8759	0.7743	0.9645	0.9189	0.9356	0.7035	0.7064	0.8856	0.9796	0.8370	0.9103	0.5719	0.5054	0.5426	0.6418	0.7274	0.9125	0.5885	0.9426	0.7510	0.5055	0.9346	0.7204	0.7279	0.6447	0.4707	0.7137
0.6038	3.2258	1200	0.5645	0.7330	0.8552	0.9526	0.9843	0.9456	0.9536	0.7353	0.7254	0.7528	0.8507	0.8675	0.9411	0.7901	0.9830	0.8741	0.7417	0.9695	0.7350	0.8992	0.8817	0.7758	0.8421	0.9799	0.8397	0.9131	0.6145	0.5277	0.5483	0.6496	0.7479	0.9124	0.5919	0.9413	0.7624	0.5452	0.9372	0.6455	0.7698	0.7256	0.5486	0.7257
0.6218	3.4946	1300	0.5641	0.7224	0.8571	0.9524	0.9863	0.9358	0.9425	0.7547	0.6836	0.7491	0.8775	0.8444	0.9505	0.8109	0.9820	0.9127	0.7426	0.9710	0.6970	0.9102	0.9148	0.7668	0.8517	0.9810	0.8483	0.9106	0.6126	0.5573	0.5396	0.6359	0.7261	0.9120	0.6001	0.9429	0.7474	0.5381	0.9369	0.6221	0.7279	0.6511	0.5116	0.7233
0.5334	3.7634	1400	0.5687	0.7141	0.8338	0.9518	0.9861	0.9403	0.9549	0.7046	0.7106	0.6880	0.8719	0.8349	0.9442	0.6876	0.9817	0.8778	0.5796	0.9746	0.7663	0.9041	0.7933	0.7614	0.8798	0.9809	0.8418	0.9125	0.5459	0.5277	0.5466	0.6398	0.7499	0.9115	0.5282	0.9396	0.7568	0.4521	0.9319	0.6160	0.7445	0.6995	0.5142	0.7285

Framework versions

• Transformers 4.48.0
• Pytorch 2.1.2+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0