segformer-b3-fashion-finetuned-polo-segments-v1.4

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: 0.0547
• Mean Iou: 0.7482
• Mean Accuracy: 0.9206
• Overall Accuracy: 0.9823
• Accuracy Unlabeled: nan
• Accuracy Collar: 0.8807
• Accuracy Polo: 0.9847
• Accuracy Lines-cuff: 0.7598
• Accuracy Lines-chest: 0.9230
• Accuracy Human: 0.9823
• Accuracy Background: 0.9929
• Accuracy Tape: nan
• Iou Unlabeled: nan
• Iou Collar: 0.8276
• Iou Polo: 0.9580
• Iou Lines-cuff: 0.6735
• Iou Lines-chest: 0.8230
• Iou Human: 0.9680
• Iou Background: 0.9872
• Iou Tape: 0.0

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
0.206	2.5	20	0.1808	0.5772	0.6083	0.9552	nan	0.6933	0.9875	0.0	0.0304	0.9818	0.9566	nan	nan	0.6406	0.9100	0.0	0.0276	0.9299	0.9553	nan
0.0873	5.0	40	0.0882	0.7806	0.8207	0.9768	nan	0.8457	0.9848	0.2359	0.8896	0.9780	0.9904	nan	nan	0.7808	0.9460	0.2351	0.7783	0.9605	0.9827	nan
0.0648	7.5	60	0.0712	0.8502	0.8900	0.9794	nan	0.8586	0.9880	0.6659	0.8584	0.9796	0.9892	nan	nan	0.8059	0.9499	0.6054	0.7918	0.9642	0.9842	nan
0.0607	10.0	80	0.0631	0.8556	0.8957	0.9806	nan	0.8586	0.9856	0.7087	0.8477	0.9829	0.9907	nan	nan	0.8055	0.9539	0.6394	0.7834	0.9659	0.9856	nan
0.057	12.5	100	0.0605	0.8661	0.9135	0.9815	nan	0.8708	0.9818	0.7296	0.9224	0.9855	0.9908	nan	nan	0.8148	0.9570	0.6577	0.8144	0.9669	0.9859	nan
0.0458	15.0	120	0.0573	0.7446	0.9169	0.9819	nan	0.8925	0.9838	0.7505	0.9009	0.9792	0.9949	nan	nan	0.8244	0.9581	0.6600	0.8164	0.9669	0.9863	0.0
0.0413	17.5	140	0.0587	0.7428	0.9196	0.9818	nan	0.8818	0.9820	0.7483	0.9299	0.9823	0.9932	nan	nan	0.8217	0.9571	0.6671	0.7997	0.9673	0.9869	0.0
0.0449	20.0	160	0.0542	0.7468	0.9202	0.9826	nan	0.8850	0.9833	0.7516	0.9248	0.9842	0.9925	nan	nan	0.8270	0.9590	0.6678	0.8179	0.9688	0.9873	0.0
0.0394	22.5	180	0.0558	0.7468	0.9208	0.9819	nan	0.8934	0.9853	0.7528	0.9207	0.9808	0.9919	nan	nan	0.8298	0.9564	0.6657	0.8214	0.9672	0.9869	0.0
0.0472	25.0	200	0.0549	0.7474	0.9185	0.9823	nan	0.8792	0.9854	0.7531	0.9186	0.9828	0.9922	nan	nan	0.8274	0.9577	0.6681	0.8233	0.9681	0.9871	0.0
0.0452	27.5	220	0.0547	0.7482	0.9217	0.9823	nan	0.8837	0.9846	0.7622	0.9247	0.9823	0.9927	nan	nan	0.8287	0.9580	0.6733	0.8221	0.9681	0.9871	0.0
0.0392	30.0	240	0.0547	0.7482	0.9206	0.9823	nan	0.8807	0.9847	0.7598	0.9230	0.9823	0.9929	nan	nan	0.8276	0.9580	0.6735	0.8230	0.9680	0.9872	0.0

Framework versions

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