segformer-b0-finetuned-segments-sidewalk-outputs

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

• Loss: 0.0650
• Mean Iou: 0.5841
• Mean Accuracy: 0.8778
• Overall Accuracy: 0.9588
• Accuracy Background: nan
• Accuracy Ground: 0.9811
• Accuracy Pallet: 0.7744
• Iou Background: 0.0
• Iou Ground: 0.9805
• Iou Pallet: 0.7717

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Ground	Accuracy Pallet	Iou Background	Iou Ground	Iou Pallet
0.0421	0.1471	20	0.0709	0.5844	0.8798	0.9604	nan	0.9826	0.7771	0.0	0.9808	0.7724
0.0276	0.2941	40	0.0624	0.6015	0.9042	0.9646	nan	0.9813	0.8271	0.0	0.9809	0.8236
0.0211	0.4412	60	0.0691	0.5664	0.8509	0.9552	nan	0.9840	0.7179	0.0	0.9835	0.7157
0.0235	0.5882	80	0.0621	0.5617	0.8437	0.9565	nan	0.9876	0.6998	0.0	0.9866	0.6984
0.0271	0.7353	100	0.0602	0.5877	0.8832	0.9567	nan	0.9770	0.7893	0.0	0.9767	0.7863
0.024	0.8824	120	0.0682	0.5923	0.8911	0.9600	nan	0.9790	0.8033	0.0	0.9785	0.7985
0.0407	1.0294	140	0.0691	0.5892	0.8859	0.9580	nan	0.9779	0.7940	0.0	0.9773	0.7903
0.0277	1.1765	160	0.0683	0.5788	0.8697	0.9571	nan	0.9812	0.7582	0.0	0.9804	0.7560
0.017	1.3235	180	0.0679	0.5845	0.8789	0.9566	nan	0.9780	0.7798	0.0	0.9770	0.7764
0.0275	1.4706	200	0.0634	0.5992	0.9014	0.9639	nan	0.9812	0.8216	0.0	0.9800	0.8175
0.0122	1.6176	220	0.0602	0.5859	0.8813	0.9590	nan	0.9804	0.7821	0.0	0.9794	0.7782
0.0149	1.7647	240	0.0662	0.5827	0.8757	0.9591	nan	0.9820	0.7694	0.0	0.9817	0.7664
0.0169	1.9118	260	0.0628	0.5994	0.9019	0.9614	nan	0.9778	0.8259	0.0	0.9776	0.8205
0.0324	2.0588	280	0.0677	0.5859	0.8809	0.9584	nan	0.9798	0.7819	0.0	0.9792	0.7785
0.0229	2.2059	300	0.0693	0.5983	0.9003	0.9619	nan	0.9789	0.8217	0.0	0.9784	0.8166
0.0204	2.3529	320	0.0729	0.5850	0.8792	0.9586	nan	0.9805	0.7780	0.0	0.9800	0.7749
0.0102	2.5	340	0.0655	0.5899	0.8868	0.9603	nan	0.9806	0.7931	0.0	0.9802	0.7895
0.0235	2.6471	360	0.0682	0.5781	0.8688	0.9556	nan	0.9795	0.7580	0.0	0.9789	0.7552
0.0239	2.7941	380	0.0633	0.5961	0.8965	0.9615	nan	0.9794	0.8136	0.0	0.9789	0.8093
0.0305	2.9412	400	0.0593	0.5832	0.8764	0.9593	nan	0.9822	0.7706	0.0	0.9817	0.7678
0.0183	3.0882	420	0.0600	0.5867	0.8816	0.9613	nan	0.9832	0.7799	0.0	0.9827	0.7775
0.031	3.2353	440	0.0612	0.5933	0.8917	0.9614	nan	0.9806	0.8029	0.0	0.9799	0.8000
0.0174	3.3824	460	0.0645	0.5836	0.8769	0.9590	nan	0.9816	0.7722	0.0	0.9811	0.7696
0.0456	3.5294	480	0.0651	0.5770	0.8669	0.9577	nan	0.9827	0.7512	0.0	0.9821	0.7489
0.0187	3.6765	500	0.0659	0.5831	0.8765	0.9578	nan	0.9803	0.7727	0.0	0.9798	0.7695
0.0329	3.8235	520	0.0690	0.5787	0.8697	0.9568	nan	0.9808	0.7587	0.0	0.9801	0.7560
0.0241	3.9706	540	0.0651	0.5847	0.8789	0.9584	nan	0.9803	0.7774	0.0	0.9798	0.7743
0.0304	4.1176	560	0.0652	0.5871	0.8823	0.9589	nan	0.9800	0.7846	0.0	0.9795	0.7817
0.0086	4.2647	580	0.0662	0.5851	0.8793	0.9584	nan	0.9802	0.7784	0.0	0.9797	0.7756
0.0194	4.4118	600	0.0678	0.5889	0.8853	0.9581	nan	0.9781	0.7925	0.0	0.9777	0.7890
0.0114	4.5588	620	0.0664	0.5877	0.8834	0.9582	nan	0.9789	0.7880	0.0	0.9784	0.7847
0.0183	4.7059	640	0.0663	0.5843	0.8782	0.9571	nan	0.9789	0.7776	0.0	0.9784	0.7744
0.0139	4.8529	660	0.0652	0.5872	0.8826	0.9590	nan	0.9801	0.7852	0.0	0.9796	0.7820
0.0197	5.0	680	0.0650	0.5841	0.8778	0.9588	nan	0.9811	0.7744	0.0	0.9805	0.7717

Framework versions

• Transformers 4.46.3
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3