dropoff-utcustom-train-SF-RGB-b5_4

This model is a fine-tuned version of nvidia/mit-b5 on the sam1120/dropoff-utcustom-TRAIN dataset. It achieves the following results on the evaluation set:

• Loss: 0.2242
• Mean Iou: 0.4568
• Mean Accuracy: 0.7402
• Overall Accuracy: 0.9696
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.4899
• Accuracy Undropoff: 0.9904
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.4016
• Iou Undropoff: 0.9690

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: 7e-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 120

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Dropoff	Accuracy Undropoff	Iou Unlabeled	Iou Dropoff	Iou Undropoff
0.9465	5.0	10	0.9974	0.2695	0.5001	0.6771	nan	0.3071	0.6931	0.0	0.1261	0.6824
0.8558	10.0	20	0.8237	0.3822	0.7119	0.8664	nan	0.5434	0.8804	0.0	0.2787	0.8678
0.7585	15.0	30	0.6801	0.4232	0.7487	0.9194	nan	0.5625	0.9349	0.0	0.3494	0.9202
0.715	20.0	40	0.6076	0.4298	0.7663	0.9232	nan	0.5952	0.9375	0.0	0.3661	0.9233
0.6145	25.0	50	0.5298	0.4398	0.7760	0.9380	nan	0.5994	0.9527	0.0	0.3819	0.9375
0.5355	30.0	60	0.4821	0.4426	0.7749	0.9428	nan	0.5918	0.9581	0.0	0.3857	0.9422
0.4619	35.0	70	0.4266	0.4493	0.7716	0.9524	nan	0.5743	0.9688	0.0	0.3962	0.9517
0.4367	40.0	80	0.3941	0.4519	0.7738	0.9568	nan	0.5742	0.9734	0.0	0.3997	0.9559
0.3839	45.0	90	0.3801	0.4528	0.7796	0.9577	nan	0.5853	0.9738	0.0	0.4017	0.9567
0.3164	50.0	100	0.3549	0.4543	0.7785	0.9608	nan	0.5797	0.9773	0.0	0.4030	0.9599
0.3018	55.0	110	0.3327	0.4573	0.7731	0.9639	nan	0.5650	0.9812	0.0	0.4087	0.9631
0.2646	60.0	120	0.3127	0.4590	0.7703	0.9658	nan	0.5571	0.9835	0.0	0.4121	0.9650
0.2378	65.0	130	0.2958	0.4628	0.7728	0.9673	nan	0.5607	0.9850	0.0	0.4217	0.9666
0.2076	70.0	140	0.2778	0.4675	0.7729	0.9693	nan	0.5586	0.9871	0.0	0.4340	0.9686
0.1951	75.0	150	0.2648	0.4666	0.7719	0.9692	nan	0.5567	0.9871	0.0	0.4314	0.9685
0.1734	80.0	160	0.2522	0.4673	0.7643	0.9703	nan	0.5397	0.9890	0.0	0.4322	0.9696
0.1569	85.0	170	0.2436	0.4660	0.7603	0.9703	nan	0.5312	0.9894	0.0	0.4282	0.9697
0.1691	90.0	180	0.2411	0.4647	0.7624	0.9697	nan	0.5363	0.9885	0.0	0.4250	0.9690
0.1498	95.0	190	0.2335	0.4623	0.7537	0.9699	nan	0.5179	0.9895	0.0	0.4176	0.9692
0.1478	100.0	200	0.2281	0.4585	0.7420	0.9700	nan	0.4934	0.9906	0.0	0.4062	0.9693
0.1407	105.0	210	0.2278	0.4615	0.7501	0.9701	nan	0.5102	0.9900	0.0	0.4151	0.9694
0.1397	110.0	220	0.2305	0.4610	0.7512	0.9698	nan	0.5129	0.9896	0.0	0.4140	0.9691
0.1317	115.0	230	0.2265	0.4576	0.7430	0.9695	nan	0.4959	0.9901	0.0	0.4038	0.9689
0.1548	120.0	240	0.2242	0.4568	0.7402	0.9696	nan	0.4899	0.9904	0.0	0.4016	0.9690

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu117
• Datasets 2.13.1
• Tokenizers 0.13.3