segformer-b1-finetuned-cityscapes-1024-1024-with-after-demo-ds

This model is a fine-tuned version of nvidia/segformer-b1-finetuned-cityscapes-1024-1024 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0153
• Mean Iou: 0.9689
• Mean Accuracy: 0.9858
• Overall Accuracy: 0.9947
• Accuracy Default: 1e-06
• Accuracy Pipe: 0.9729
• Accuracy Floor: 0.9861
• Accuracy Background: 0.9985
• Iou Default: 1e-06
• Iou Pipe: 0.9305
• Iou Floor: 0.9802
• Iou Background: 0.9958

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 60

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Default	Accuracy Pipe	Accuracy Floor	Accuracy Background	Iou Default	Iou Pipe	Iou Floor	Iou Background
0.333	1.0	55	0.1193	0.8358	0.8688	0.9725	1e-06	0.6617	0.9467	0.9981	1e-06	0.5954	0.9420	0.9700
0.0978	2.0	110	0.0734	0.8938	0.9399	0.9817	1e-06	0.8567	0.9709	0.9921	1e-06	0.7472	0.9523	0.9818
0.0647	3.0	165	0.0529	0.9169	0.9580	0.9860	1e-06	0.9093	0.9696	0.9951	1e-06	0.8023	0.9617	0.9866
0.0519	4.0	220	0.0455	0.9175	0.9445	0.9861	1e-06	0.8663	0.9692	0.9979	1e-06	0.8031	0.9638	0.9855
0.0457	5.0	275	0.0413	0.9198	0.9687	0.9866	1e-06	0.9356	0.9786	0.9919	1e-06	0.8098	0.9614	0.9881
0.0407	6.0	330	0.0360	0.9283	0.9584	0.9882	1e-06	0.9010	0.9780	0.9962	1e-06	0.8320	0.9632	0.9897
0.0363	7.0	385	0.0318	0.9399	0.9698	0.9897	1e-06	0.9385	0.9737	0.9973	1e-06	0.8614	0.9680	0.9904
0.0335	8.0	440	0.0295	0.9423	0.9727	0.9904	1e-06	0.9443	0.9770	0.9969	1e-06	0.8652	0.9702	0.9915
0.0318	9.0	495	0.0288	0.9425	0.9746	0.9905	1e-06	0.9492	0.9784	0.9963	1e-06	0.8664	0.9694	0.9918
0.0292	10.0	550	0.0262	0.9478	0.9752	0.9912	1e-06	0.9510	0.9769	0.9976	1e-06	0.8803	0.9710	0.9922
0.0291	11.0	605	0.0270	0.9466	0.9720	0.9909	1e-06	0.9415	0.9765	0.9979	1e-06	0.8774	0.9708	0.9916
0.0275	12.0	660	0.0249	0.9496	0.9793	0.9916	1e-06	0.9625	0.9784	0.9971	1e-06	0.8835	0.9723	0.9929
0.0264	13.0	715	0.0246	0.9514	0.9716	0.9915	1e-06	0.9383	0.9782	0.9984	1e-06	0.8901	0.9720	0.9920
0.0255	14.0	770	0.0242	0.9500	0.9812	0.9917	1e-06	0.9677	0.9792	0.9967	1e-06	0.8846	0.9723	0.9932
0.0248	15.0	825	0.0230	0.9534	0.9785	0.9921	1e-06	0.9598	0.9777	0.9980	1e-06	0.8940	0.9732	0.9931
0.0241	16.0	880	0.0233	0.9523	0.9806	0.9920	1e-06	0.9666	0.9778	0.9975	1e-06	0.8906	0.9731	0.9932
0.023	17.0	935	0.0215	0.9562	0.9778	0.9925	1e-06	0.9553	0.9801	0.9982	1e-06	0.9015	0.9738	0.9934
0.0223	18.0	990	0.0212	0.9562	0.9780	0.9925	1e-06	0.9546	0.9816	0.9979	1e-06	0.9011	0.9737	0.9937
0.022	19.0	1045	0.0205	0.9558	0.9810	0.9927	1e-06	0.9640	0.9813	0.9975	1e-06	0.8995	0.9737	0.9941
0.0213	20.0	1100	0.0207	0.9582	0.9764	0.9926	1e-06	0.9504	0.9801	0.9986	1e-06	0.9069	0.9745	0.9932
0.0213	21.0	1155	0.0211	0.9566	0.9801	0.9927	1e-06	0.9624	0.9796	0.9981	1e-06	0.9014	0.9746	0.9937
0.0206	22.0	1210	0.0202	0.9589	0.9799	0.9929	1e-06	0.9608	0.9804	0.9983	1e-06	0.9078	0.9752	0.9938
0.0199	23.0	1265	0.0194	0.9596	0.9813	0.9931	1e-06	0.9644	0.9812	0.9981	1e-06	0.9096	0.9750	0.9942
0.0192	24.0	1320	0.0194	0.9590	0.9831	0.9932	1e-06	0.9710	0.9803	0.9981	1e-06	0.9070	0.9754	0.9945
0.019	25.0	1375	0.0189	0.9608	0.9834	0.9933	1e-06	0.9703	0.9820	0.9978	1e-06	0.9124	0.9754	0.9945
0.0189	26.0	1430	0.0195	0.9602	0.9822	0.9932	1e-06	0.9675	0.9808	0.9983	1e-06	0.9103	0.9758	0.9943
0.0185	27.0	1485	0.0204	0.9577	0.9804	0.9930	1e-06	0.9617	0.9815	0.9981	1e-06	0.9035	0.9754	0.9942
0.0185	28.0	1540	0.0188	0.9625	0.9808	0.9935	1e-06	0.9616	0.9822	0.9986	1e-06	0.9167	0.9766	0.9944
0.0178	29.0	1595	0.0186	0.9626	0.9801	0.9935	1e-06	0.9588	0.9829	0.9985	1e-06	0.9166	0.9768	0.9943
0.0176	30.0	1650	0.0192	0.9622	0.9802	0.9935	1e-06	0.9594	0.9826	0.9986	1e-06	0.9156	0.9766	0.9945
0.0175	31.0	1705	0.0175	0.9631	0.9839	0.9937	1e-06	0.9710	0.9827	0.9981	1e-06	0.9176	0.9769	0.9948
0.017	32.0	1760	0.0183	0.9615	0.9852	0.9936	1e-06	0.9761	0.9814	0.9981	1e-06	0.9130	0.9765	0.9949
0.0172	33.0	1815	0.0173	0.9646	0.9834	0.9938	1e-06	0.9690	0.9830	0.9984	1e-06	0.9218	0.9772	0.9948
0.0167	34.0	1870	0.0175	0.9625	0.9857	0.9938	1e-06	0.9768	0.9822	0.9981	1e-06	0.9156	0.9769	0.9951
0.0164	35.0	1925	0.0170	0.9643	0.9854	0.9940	1e-06	0.9749	0.9832	0.9981	1e-06	0.9200	0.9776	0.9952
0.016	36.0	1980	0.0166	0.9657	0.9844	0.9941	1e-06	0.9710	0.9837	0.9984	1e-06	0.9237	0.9782	0.9952
0.0161	37.0	2035	0.0169	0.9661	0.9830	0.9941	1e-06	0.9668	0.9834	0.9987	1e-06	0.9254	0.9780	0.9949
0.0156	38.0	2090	0.0172	0.9648	0.9840	0.9939	1e-06	0.9706	0.9829	0.9984	1e-06	0.9220	0.9774	0.9949
0.0156	39.0	2145	0.0170	0.9640	0.9857	0.9940	1e-06	0.9769	0.9817	0.9985	1e-06	0.9192	0.9774	0.9953
0.0152	40.0	2200	0.0164	0.9667	0.9845	0.9942	1e-06	0.9710	0.9839	0.9985	1e-06	0.9267	0.9783	0.9952
0.0153	41.0	2255	0.0164	0.9663	0.9854	0.9942	1e-06	0.9748	0.9830	0.9985	1e-06	0.9256	0.9780	0.9953
0.016	42.0	2310	0.0162	0.9662	0.9854	0.9942	1e-06	0.9744	0.9833	0.9985	1e-06	0.9254	0.9778	0.9954
0.0157	43.0	2365	0.0162	0.9670	0.9849	0.9943	1e-06	0.9724	0.9837	0.9986	1e-06	0.9269	0.9786	0.9953
0.0148	44.0	2420	0.0167	0.9671	0.9850	0.9943	1e-06	0.9719	0.9849	0.9983	1e-06	0.9273	0.9786	0.9953
0.0149	45.0	2475	0.0165	0.9660	0.9853	0.9943	1e-06	0.9730	0.9846	0.9983	1e-06	0.9235	0.9789	0.9955
0.0144	46.0	2530	0.0154	0.9670	0.9870	0.9945	1e-06	0.9784	0.9844	0.9983	1e-06	0.9260	0.9791	0.9958
0.0142	47.0	2585	0.0150	0.9685	0.9865	0.9946	1e-06	0.9762	0.9847	0.9985	1e-06	0.9302	0.9794	0.9957
0.0142	48.0	2640	0.0154	0.9672	0.9870	0.9945	1e-06	0.9784	0.9841	0.9984	1e-06	0.9268	0.9792	0.9957
0.0144	49.0	2695	0.0152	0.9677	0.9862	0.9945	1e-06	0.9754	0.9847	0.9985	1e-06	0.9284	0.9791	0.9957
0.0141	50.0	2750	0.0154	0.9681	0.9857	0.9946	1e-06	0.9729	0.9857	0.9984	1e-06	0.9289	0.9796	0.9957
0.0136	51.0	2805	0.0153	0.9690	0.9855	0.9947	1e-06	0.9728	0.9850	0.9987	1e-06	0.9317	0.9797	0.9957
0.0138	52.0	2860	0.0150	0.9691	0.9866	0.9947	1e-06	0.9767	0.9846	0.9986	1e-06	0.9320	0.9796	0.9957
0.014	53.0	2915	0.0158	0.9673	0.9853	0.9945	1e-06	0.9720	0.9855	0.9984	1e-06	0.9266	0.9798	0.9956
0.0136	54.0	2970	0.0154	0.9693	0.9857	0.9948	1e-06	0.9725	0.9863	0.9985	1e-06	0.9319	0.9802	0.9958
0.0138	55.0	3025	0.0154	0.9692	0.9853	0.9947	1e-06	0.9717	0.9855	0.9986	1e-06	0.9323	0.9798	0.9956
0.0134	56.0	3080	0.0153	0.9689	0.9857	0.9947	1e-06	0.9728	0.9860	0.9984	1e-06	0.9312	0.9797	0.9957
0.0135	57.0	3135	0.0154	0.9695	0.9863	0.9948	1e-06	0.9747	0.9855	0.9986	1e-06	0.9325	0.9800	0.9958
0.0133	58.0	3190	0.0154	0.9689	0.9859	0.9947	1e-06	0.9739	0.9854	0.9985	1e-06	0.9313	0.9798	0.9957
0.0134	59.0	3245	0.0152	0.9696	0.9862	0.9948	1e-06	0.9745	0.9856	0.9986	1e-06	0.9328	0.9801	0.9958
0.0138	60.0	3300	0.0153	0.9689	0.9858	0.9947	1e-06	0.9729	0.9861	0.9985	1e-06	0.9305	0.9802	0.9958

Framework versions

• Transformers 4.35.2
• Pytorch 2.0.1
• Datasets 2.15.0
• Tokenizers 0.15.0