segformer-b0-finetuned-segments-ECHO-dev-01-v1

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

• Loss: 0.1147
• Mean Iou: 0.7718
• Mean Accuracy: 0.9647
• Overall Accuracy: 0.9651
• Accuracy Unlabeled: nan
• Accuracy Lv: 0.9592
• Accuracy Lr: 0.9722
• Accuracy Ra: 0.9540
• Accuracy La: 0.9733
• Iou Unlabeled: 0.0
• Iou Lv: 0.9592
• Iou Lr: 0.9722
• Iou Ra: 0.9540
• Iou La: 0.9733

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Lv	Accuracy Lr	Accuracy Ra	Accuracy La	Iou Unlabeled	Iou Lv	Iou Lr	Iou Ra	Iou La
0.9325	2.5	20	1.3095	0.4524	0.7068	0.7433	nan	0.9170	0.8666	0.2895	0.7542	0.0	0.7498	0.6467	0.2538	0.6116
0.6633	5.0	40	0.7431	0.6359	0.8411	0.8600	nan	0.9559	0.9196	0.6304	0.8583	0.0	0.8810	0.8526	0.6270	0.8191
0.5492	7.5	60	0.4682	0.7583	0.9515	0.9548	nan	0.9702	0.9610	0.9452	0.9295	0.0	0.9630	0.9610	0.9380	0.9295
0.5418	10.0	80	0.3925	0.7453	0.9323	0.9367	nan	0.9520	0.9515	0.9050	0.9206	0.0	0.9509	0.9515	0.9036	0.9206
0.3586	12.5	100	0.3216	0.7612	0.9515	0.9561	nan	0.9594	0.9767	0.9359	0.9340	0.0	0.9594	0.9767	0.9359	0.9340
0.2943	15.0	120	0.2719	0.7615	0.9518	0.9541	nan	0.9480	0.9691	0.9413	0.9490	0.0	0.9480	0.9691	0.9413	0.9490
0.3039	17.5	140	0.2505	0.7765	0.9763	0.9743	nan	0.9742	0.9650	0.9799	0.9861	0.0	0.9742	0.9650	0.9572	0.9861
0.2639	20.0	160	0.2208	0.7743	0.9679	0.9690	nan	0.9676	0.9784	0.9477	0.9778	0.0	0.9676	0.9784	0.9477	0.9778
0.2252	22.5	180	0.1928	0.7710	0.9637	0.9655	nan	0.9618	0.9784	0.9483	0.9664	0.0	0.9618	0.9784	0.9483	0.9664
0.1816	25.0	200	0.1756	0.7690	0.9612	0.9619	nan	0.9574	0.9696	0.9501	0.9679	0.0	0.9574	0.9696	0.9501	0.9679
0.1676	27.5	220	0.1556	0.7610	0.9513	0.9541	nan	0.9483	0.9721	0.9387	0.9460	0.0	0.9483	0.9721	0.9387	0.9460
0.1833	30.0	240	0.1468	0.7786	0.9733	0.9742	nan	0.9669	0.9837	0.9639	0.9786	0.0	0.9669	0.9837	0.9639	0.9786
0.1487	32.5	260	0.1367	0.7708	0.9635	0.9649	nan	0.9580	0.9776	0.9479	0.9705	0.0	0.9580	0.9776	0.9479	0.9705
0.1482	35.0	280	0.1320	0.7712	0.9641	0.9655	nan	0.9576	0.9779	0.9555	0.9653	0.0	0.9576	0.9779	0.9555	0.9653
0.1412	37.5	300	0.1241	0.7763	0.9704	0.9706	nan	0.9696	0.9738	0.9590	0.9791	0.0	0.9696	0.9738	0.9590	0.9791
0.1282	40.0	320	0.1213	0.7724	0.9655	0.9665	nan	0.9588	0.9778	0.9512	0.9742	0.0	0.9588	0.9778	0.9512	0.9742
0.133	42.5	340	0.1155	0.7745	0.9681	0.9686	nan	0.9640	0.9752	0.9580	0.9751	0.0	0.9640	0.9752	0.9580	0.9751
0.1172	45.0	360	0.1178	0.7673	0.9603	0.9607	nan	0.9562	0.9665	0.9510	0.9676	0.0	0.9562	0.9665	0.9460	0.9676
0.1469	47.5	380	0.1138	0.7734	0.9668	0.9673	nan	0.9629	0.9743	0.9545	0.9753	0.0	0.9629	0.9743	0.9545	0.9753
0.1247	50.0	400	0.1147	0.7718	0.9647	0.9651	nan	0.9592	0.9722	0.9540	0.9733	0.0	0.9592	0.9722	0.9540	0.9733

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0