dropoff-utcustom-train-SF-RGB-b5_2

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: 0.4848
• Mean Iou: 0.4257
• Mean Accuracy: 0.7972
• Overall Accuracy: 0.9466
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.6343
• Accuracy Undropoff: 0.9601
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.3321
• Iou Undropoff: 0.9451

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: 3e-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
1.0108	5.0	10	1.0721	0.1514	0.5401	0.4205	nan	0.6706	0.4096	0.0	0.0494	0.4047
0.9654	10.0	20	0.9802	0.2190	0.6570	0.5944	nan	0.7253	0.5887	0.0	0.0745	0.5826
0.9175	15.0	30	0.9047	0.2553	0.7350	0.6792	nan	0.7960	0.6741	0.0	0.0973	0.6686
0.9052	20.0	40	0.8427	0.2812	0.7661	0.7377	nan	0.7971	0.7351	0.0	0.1146	0.7290
0.8555	25.0	50	0.7970	0.3063	0.7827	0.7900	nan	0.7748	0.7906	0.0	0.1357	0.7832
0.8291	30.0	60	0.7543	0.3289	0.7891	0.8332	nan	0.7410	0.8372	0.0	0.1586	0.8282
0.7923	35.0	70	0.7327	0.3375	0.7961	0.8471	nan	0.7405	0.8517	0.0	0.1701	0.8425
0.7724	40.0	80	0.6994	0.3529	0.7968	0.8719	nan	0.7149	0.8787	0.0	0.1906	0.8682
0.7215	45.0	90	0.6675	0.3694	0.7935	0.8954	nan	0.6824	0.9047	0.0	0.2157	0.8926
0.6907	50.0	100	0.6521	0.3742	0.7998	0.9000	nan	0.6904	0.9091	0.0	0.2252	0.8973
0.6768	55.0	110	0.6260	0.3850	0.8022	0.9118	nan	0.6827	0.9217	0.0	0.2455	0.9094
0.659	60.0	120	0.6010	0.3965	0.7973	0.9244	nan	0.6586	0.9359	0.0	0.2671	0.9224
0.6265	65.0	130	0.5847	0.4005	0.7992	0.9276	nan	0.6592	0.9393	0.0	0.2757	0.9258
0.6134	70.0	140	0.5673	0.4060	0.8022	0.9316	nan	0.6611	0.9433	0.0	0.2881	0.9297
0.5864	75.0	150	0.5401	0.4132	0.7961	0.9383	nan	0.6410	0.9511	0.0	0.3029	0.9366
0.5686	80.0	160	0.5289	0.4153	0.7974	0.9395	nan	0.6424	0.9524	0.0	0.3080	0.9379
0.5597	85.0	170	0.5386	0.4114	0.8079	0.9350	nan	0.6692	0.9465	0.0	0.3011	0.9331
0.5718	90.0	180	0.5080	0.4210	0.7947	0.9438	nan	0.6321	0.9573	0.0	0.3208	0.9423
0.517	95.0	190	0.5026	0.4222	0.7956	0.9445	nan	0.6332	0.9580	0.0	0.3236	0.9430
0.5252	100.0	200	0.4990	0.4232	0.7969	0.9450	nan	0.6354	0.9584	0.0	0.3261	0.9435
0.5174	105.0	210	0.4951	0.4223	0.8012	0.9437	nan	0.6457	0.9567	0.0	0.3249	0.9422
0.5217	110.0	220	0.4882	0.4238	0.7993	0.9450	nan	0.6404	0.9582	0.0	0.3280	0.9435
0.5224	115.0	230	0.4846	0.4258	0.7968	0.9467	nan	0.6333	0.9603	0.0	0.3321	0.9452
0.5399	120.0	240	0.4848	0.4257	0.7972	0.9466	nan	0.6343	0.9601	0.0	0.3321	0.9451

Framework versions

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