dropoff-utcustom-train-SF-RGB-b5_6

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: 0.2315
• Mean Iou: 0.6980
• Mean Accuracy: 0.7503
• Overall Accuracy: 0.9714
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.5091
• Accuracy Undropoff: 0.9915
• Iou Unlabeled: nan
• Iou Dropoff: 0.4253
• Iou Undropoff: 0.9708

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: 2e-05
• 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
1.0694	5.0	10	1.0190	0.2533	0.6371	0.6676	nan	0.6038	0.6703	0.0	0.0976	0.6624
0.8457	10.0	20	0.7681	0.4126	0.7662	0.9307	nan	0.5867	0.9457	0.0	0.3078	0.9300
0.6049	15.0	30	0.5718	0.4362	0.7527	0.9568	nan	0.5301	0.9753	0.0	0.3527	0.9561
0.5206	20.0	40	0.4181	0.4522	0.7468	0.9662	nan	0.5076	0.9861	0.0	0.3909	0.9656
0.3478	25.0	50	0.3144	0.4603	0.7376	0.9709	nan	0.4832	0.9920	0.0	0.4105	0.9705
0.2023	30.0	60	0.2893	0.4654	0.7612	0.9701	nan	0.5332	0.9891	0.0	0.4267	0.9695
0.1367	35.0	70	0.2351	0.6813	0.7176	0.9715	nan	0.4407	0.9946	nan	0.3916	0.9710
0.1272	40.0	80	0.2364	0.6824	0.7217	0.9713	nan	0.4495	0.9939	nan	0.3941	0.9707
0.0929	45.0	90	0.2536	0.4704	0.7617	0.9718	nan	0.5326	0.9909	0.0	0.4401	0.9712
0.0756	50.0	100	0.2253	0.6950	0.7479	0.9710	nan	0.5045	0.9912	nan	0.4197	0.9704
0.0756	55.0	110	0.2305	0.7043	0.7606	0.9716	nan	0.5305	0.9908	nan	0.4375	0.9710
0.0721	60.0	120	0.2213	0.6964	0.7448	0.9716	nan	0.4974	0.9922	nan	0.4218	0.9711
0.0683	65.0	130	0.2338	0.7047	0.7631	0.9715	nan	0.5359	0.9904	nan	0.4385	0.9708
0.0642	70.0	140	0.2314	0.7046	0.7637	0.9714	nan	0.5373	0.9902	nan	0.4385	0.9707
0.0623	75.0	150	0.2205	0.7013	0.7565	0.9714	nan	0.5222	0.9909	nan	0.4317	0.9708
0.0601	80.0	160	0.2209	0.6983	0.7496	0.9715	nan	0.5075	0.9917	nan	0.4257	0.9709
0.0557	85.0	170	0.2067	0.6982	0.7463	0.9719	nan	0.5003	0.9923	nan	0.4252	0.9713
0.0571	90.0	180	0.2354	0.7022	0.7603	0.9712	nan	0.5302	0.9904	nan	0.4339	0.9706
0.0544	95.0	190	0.2240	0.7010	0.7562	0.9714	nan	0.5215	0.9909	nan	0.4311	0.9708
0.0553	100.0	200	0.2204	0.6968	0.7454	0.9717	nan	0.4987	0.9922	nan	0.4225	0.9711
0.0525	105.0	210	0.2332	0.7050	0.7625	0.9716	nan	0.5344	0.9906	nan	0.4390	0.9710
0.0524	110.0	220	0.2371	0.7033	0.7605	0.9715	nan	0.5304	0.9906	nan	0.4359	0.9708
0.0513	115.0	230	0.2333	0.6987	0.7519	0.9714	nan	0.5125	0.9913	nan	0.4267	0.9707
0.0537	120.0	240	0.2315	0.6980	0.7503	0.9714	nan	0.5091	0.9915	nan	0.4253	0.9708

Framework versions

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