segformer-b4-finetuned-segments-torso

This model is a fine-tuned version of nvidia/mit-b4 on the carllau999/torso_mask dataset. It achieves the following results on the evaluation set:

• Loss: 0.0823
• Mean Iou: 0.4815
• Mean Accuracy: 0.9630
• Overall Accuracy: 0.9630
• Accuracy Unlabeled: nan
• Accuracy Torso: 0.9630
• Iou Unlabeled: 0.0
• Iou Torso: 0.9630

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 Torso	Iou Unlabeled	Iou Torso
0.1599	1.0	20	0.1789	0.3774	0.7549	0.7549	nan	0.7549	0.0	0.7549
0.0432	2.0	40	0.1167	0.4549	0.9099	0.9099	nan	0.9099	0.0	0.9099
0.0329	3.0	60	0.1006	0.4489	0.8978	0.8978	nan	0.8978	0.0	0.8978
0.0392	4.0	80	0.1742	0.4862	0.9724	0.9724	nan	0.9724	0.0	0.9724
0.025	5.0	100	0.1047	0.4726	0.9452	0.9452	nan	0.9452	0.0	0.9452
0.0229	6.0	120	0.1330	0.4771	0.9543	0.9543	nan	0.9543	0.0	0.9543
0.0184	7.0	140	0.1060	0.4749	0.9498	0.9498	nan	0.9498	0.0	0.9498
0.0117	8.0	160	0.1163	0.4719	0.9439	0.9439	nan	0.9439	0.0	0.9439
0.0202	9.0	180	0.0790	0.4634	0.9268	0.9268	nan	0.9268	0.0	0.9268
0.0124	10.0	200	0.1063	0.4767	0.9534	0.9534	nan	0.9534	0.0	0.9534
0.0159	11.0	220	0.0979	0.4782	0.9563	0.9563	nan	0.9563	0.0	0.9563
0.0093	12.0	240	0.0850	0.4746	0.9492	0.9492	nan	0.9492	0.0	0.9492
0.0117	13.0	260	0.0870	0.4742	0.9484	0.9484	nan	0.9484	0.0	0.9484
0.0087	14.0	280	0.1058	0.4809	0.9617	0.9617	nan	0.9617	0.0	0.9617
0.0085	15.0	300	0.0897	0.4743	0.9485	0.9485	nan	0.9485	0.0	0.9485
0.01	16.0	320	0.0813	0.4774	0.9547	0.9547	nan	0.9547	0.0	0.9547
0.0109	17.0	340	0.0980	0.4810	0.9621	0.9621	nan	0.9621	0.0	0.9621
0.0076	18.0	360	0.0923	0.4787	0.9574	0.9574	nan	0.9574	0.0	0.9574
0.0067	19.0	380	0.0874	0.4778	0.9556	0.9556	nan	0.9556	0.0	0.9556
0.0079	20.0	400	0.0821	0.4766	0.9533	0.9533	nan	0.9533	0.0	0.9533
0.007	21.0	420	0.0876	0.4746	0.9491	0.9491	nan	0.9491	0.0	0.9491
0.0091	22.0	440	0.0924	0.4695	0.9390	0.9390	nan	0.9390	0.0	0.9390
0.0089	23.0	460	0.0711	0.4781	0.9563	0.9563	nan	0.9563	0.0	0.9563
0.0067	24.0	480	0.0743	0.4759	0.9518	0.9518	nan	0.9518	0.0	0.9518
0.0061	25.0	500	0.0868	0.4803	0.9605	0.9605	nan	0.9605	0.0	0.9605
0.0053	26.0	520	0.0814	0.4800	0.9599	0.9599	nan	0.9599	0.0	0.9599
0.0055	27.0	540	0.0779	0.4806	0.9612	0.9612	nan	0.9612	0.0	0.9612
0.0066	28.0	560	0.0810	0.4814	0.9628	0.9628	nan	0.9628	0.0	0.9628
0.0077	29.0	580	0.0814	0.4810	0.9621	0.9621	nan	0.9621	0.0	0.9621
0.008	30.0	600	0.0852	0.4829	0.9658	0.9658	nan	0.9658	0.0	0.9658
0.0062	31.0	620	0.0839	0.4833	0.9665	0.9665	nan	0.9665	0.0	0.9665
0.0063	32.0	640	0.0778	0.4835	0.9671	0.9671	nan	0.9671	0.0	0.9671
0.0055	33.0	660	0.0826	0.4816	0.9633	0.9633	nan	0.9633	0.0	0.9633
0.0057	34.0	680	0.0831	0.4816	0.9632	0.9632	nan	0.9632	0.0	0.9632
0.007	35.0	700	0.0809	0.4815	0.9630	0.9630	nan	0.9630	0.0	0.9630
0.007	36.0	720	0.0776	0.4809	0.9618	0.9618	nan	0.9618	0.0	0.9618
0.0075	37.0	740	0.0787	0.4809	0.9618	0.9618	nan	0.9618	0.0	0.9618
0.0054	38.0	760	0.0838	0.4817	0.9633	0.9633	nan	0.9633	0.0	0.9633
0.0061	39.0	780	0.0824	0.4820	0.9639	0.9639	nan	0.9639	0.0	0.9639
0.0063	40.0	800	0.0799	0.4825	0.9650	0.9650	nan	0.9650	0.0	0.9650
0.0065	41.0	820	0.0805	0.4813	0.9626	0.9626	nan	0.9626	0.0	0.9626
0.0052	42.0	840	0.0838	0.4822	0.9644	0.9644	nan	0.9644	0.0	0.9644
0.007	43.0	860	0.0841	0.4825	0.9650	0.9650	nan	0.9650	0.0	0.9650
0.0054	44.0	880	0.0822	0.4817	0.9634	0.9634	nan	0.9634	0.0	0.9634
0.0055	45.0	900	0.0815	0.4817	0.9633	0.9633	nan	0.9633	0.0	0.9633
0.0084	46.0	920	0.0815	0.4819	0.9638	0.9638	nan	0.9638	0.0	0.9638
0.005	47.0	940	0.0810	0.4814	0.9627	0.9627	nan	0.9627	0.0	0.9627
0.0044	48.0	960	0.0822	0.4826	0.9652	0.9652	nan	0.9652	0.0	0.9652
0.0053	49.0	980	0.0819	0.4821	0.9642	0.9642	nan	0.9642	0.0	0.9642
0.0066	50.0	1000	0.0823	0.4815	0.9630	0.9630	nan	0.9630	0.0	0.9630

Framework versions

• Transformers 4.27.4
• Pytorch 2.0.0+cpu
• Datasets 2.11.0
• Tokenizers 0.13.2