segformer-b3-fashion-finetuned-polo-segments-aug-07-v1.2

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: 0.0801
• Mean Iou: 0.8698
• Mean Accuracy: 0.9244
• Overall Accuracy: 0.9721
• Accuracy Unlabeled: nan
• Accuracy Collar: 0.8554
• Accuracy Polo: 0.9713
• Accuracy Lines-cuff: 0.8045
• Accuracy Lines-chest: 0.9543
• Accuracy Human: 0.9710
• Accuracy Background: 0.9896
• Iou Unlabeled: nan
• Iou Collar: 0.7817
• Iou Polo: 0.9387
• Iou Lines-cuff: 0.7252
• Iou Lines-chest: 0.8492
• Iou Human: 0.9456
• Iou Background: 0.9783

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	Iou Unlabeled	Iou Collar	Iou Polo	Iou Lines-cuff	Iou Lines-chest	Iou Human	Iou Background
0.1958	4.0	20	0.1935	0.5473	0.5809	0.9484	nan	0.5688	0.9710	0.0	0.0	0.9699	0.9754	nan	0.5048	0.8825	0.0	0.0	0.9291	0.9676
0.0878	8.0	40	0.1158	0.7585	0.7864	0.9643	nan	0.7548	0.9721	0.3063	0.7322	0.9787	0.9740	nan	0.7084	0.9275	0.3054	0.7048	0.9353	0.9694
0.0737	12.0	60	0.0929	0.8595	0.9043	0.9708	nan	0.8229	0.9704	0.7743	0.8982	0.9757	0.9842	nan	0.7641	0.9366	0.7031	0.8330	0.9436	0.9766
0.0646	16.0	80	0.0868	0.8643	0.9140	0.9711	nan	0.8521	0.9747	0.7778	0.9226	0.9662	0.9909	nan	0.7807	0.9359	0.7101	0.8379	0.9435	0.9774
0.0688	20.0	100	0.0819	0.8665	0.9176	0.9720	nan	0.8502	0.9721	0.7841	0.9386	0.9709	0.9899	nan	0.7814	0.9384	0.7139	0.8418	0.9459	0.9778
0.052	24.0	120	0.0821	0.8652	0.9207	0.9716	nan	0.8302	0.9646	0.8053	0.9590	0.9769	0.9883	nan	0.7694	0.9381	0.7210	0.8400	0.9445	0.9781
0.0483	28.0	140	0.0801	0.8698	0.9244	0.9721	nan	0.8554	0.9713	0.8045	0.9543	0.9710	0.9896	nan	0.7817	0.9387	0.7252	0.8492	0.9456	0.9783

Framework versions

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