segformer-b3-fashion-finetuned-polo-segments-v1.3

This model is a fine-tuned version of sayeed99/segformer-b3-fashion on the sshk/polo-badges-segmentation dataset. It achieves the following results on the evaluation set:

• Loss: 0.0429
• Mean Iou: 0.9091
• Mean Accuracy: 0.9403
• Overall Accuracy: 0.9851
• Accuracy Unlabeled: nan
• Accuracy Collar: 0.9095
• Accuracy Polo: 0.9879
• Accuracy Lines-cuff: 0.8355
• Accuracy Lines-chest: 0.9287
• Accuracy Human: 0.9883
• Accuracy Background: 0.9918
• Accuracy Tape: nan
• Iou Unlabeled: nan
• Iou Collar: 0.8597
• Iou Polo: 0.9688
• Iou Lines-cuff: 0.7831
• Iou Lines-chest: 0.8815
• Iou Human: 0.9746
• Iou Background: 0.9872
• Iou Tape: nan

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Collar	Accuracy Polo	Accuracy Lines-cuff	Accuracy Lines-chest	Accuracy Human	Accuracy Background	Accuracy Tape	Iou Unlabeled	Iou Collar	Iou Polo	Iou Lines-cuff	Iou Lines-chest	Iou Human	Iou Background	Iou Tape
0.2226	2.5	20	0.0915	0.7423	0.7696	0.9768	nan	0.8423	0.9889	0.0156	0.8004	0.9801	0.9903	nan	nan	0.8056	0.9535	0.0156	0.7379	0.9604	0.9808	nan
0.0879	5.0	40	0.0644	0.8691	0.8908	0.9806	nan	0.8701	0.9901	0.7111	0.7998	0.9908	0.9829	nan	nan	0.8372	0.9618	0.6922	0.7759	0.9674	0.9801	nan
0.0599	7.5	60	0.0525	0.8927	0.9223	0.9838	nan	0.9040	0.9855	0.7850	0.8792	0.9893	0.9911	nan	nan	0.8543	0.9668	0.7381	0.8389	0.9725	0.9855	nan
0.0517	10.0	80	0.0502	0.9011	0.9358	0.9834	nan	0.9092	0.9874	0.8282	0.9140	0.9884	0.9873	nan	nan	0.8556	0.9661	0.7672	0.8625	0.9710	0.9843	nan
0.0494	12.5	100	0.0479	0.9039	0.9372	0.9837	nan	0.9074	0.9885	0.8218	0.9300	0.9865	0.9892	nan	nan	0.8575	0.9655	0.7714	0.8721	0.9713	0.9857	nan
0.0507	15.0	120	0.0451	0.9082	0.9415	0.9844	nan	0.9126	0.9875	0.8438	0.9271	0.9869	0.9910	nan	nan	0.8592	0.9669	0.7864	0.8774	0.9728	0.9867	nan
0.0382	17.5	140	0.0460	0.9074	0.9382	0.9840	nan	0.9056	0.9897	0.8399	0.9181	0.9831	0.9930	nan	nan	0.8585	0.9651	0.7862	0.8760	0.9717	0.9870	nan
0.0365	20.0	160	0.0448	0.9104	0.9423	0.9846	nan	0.9118	0.9869	0.8552	0.9210	0.9904	0.9887	nan	nan	0.8581	0.9686	0.7969	0.8793	0.9736	0.9857	nan
0.0437	22.5	180	0.0435	0.9084	0.9397	0.9849	nan	0.9087	0.9881	0.8299	0.9323	0.9888	0.9907	nan	nan	0.8595	0.9686	0.7788	0.8824	0.9742	0.9869	nan
0.0462	25.0	200	0.0433	0.9077	0.9378	0.9850	nan	0.9076	0.9881	0.8308	0.9202	0.9886	0.9915	nan	nan	0.8597	0.9685	0.7789	0.8776	0.9743	0.9871	nan
0.0569	27.5	220	0.0428	0.9089	0.9396	0.9851	nan	0.9108	0.9879	0.8349	0.9241	0.9883	0.9917	nan	nan	0.8599	0.9688	0.7822	0.8808	0.9746	0.9872	nan
0.0357	30.0	240	0.0429	0.9091	0.9403	0.9851	nan	0.9095	0.9879	0.8355	0.9287	0.9883	0.9918	nan	nan	0.8597	0.9688	0.7831	0.8815	0.9746	0.9872	nan

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1