dropoff-utcustom-train-SF-RGBD-b5_3

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: 0.2768
• Mean Iou: 0.3194
• Mean Accuracy: 0.4999
• Overall Accuracy: 0.9578
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.0006
• Accuracy Undropoff: 0.9993
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.0006
• Iou Undropoff: 0.9578

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: 5e-06
• train_batch_size: 15
• eval_batch_size: 15
• 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
1.0992	5.0	10	1.0599	0.1938	0.4241	0.5281	nan	0.3106	0.5376	0.0	0.0540	0.5273
1.0188	10.0	20	0.9493	0.2781	0.4808	0.7846	nan	0.1494	0.8122	0.0	0.0476	0.7868
0.9218	15.0	30	0.8130	0.3074	0.4913	0.8851	nan	0.0618	0.9209	0.0	0.0364	0.8858
0.8411	20.0	40	0.7253	0.3089	0.4866	0.9038	nan	0.0315	0.9416	0.0	0.0221	0.9047
0.7583	25.0	50	0.6719	0.3097	0.4890	0.9069	nan	0.0331	0.9448	0.0	0.0216	0.9076
0.688	30.0	60	0.6303	0.3109	0.4883	0.9170	nan	0.0207	0.9559	0.0	0.0149	0.9179
0.6279	35.0	70	0.5919	0.3139	0.4918	0.9276	nan	0.0164	0.9671	0.0	0.0133	0.9283
0.5533	40.0	80	0.5375	0.3168	0.4961	0.9377	nan	0.0144	0.9777	0.0	0.0125	0.9380
0.5116	45.0	90	0.5111	0.3176	0.4970	0.9442	nan	0.0093	0.9847	0.0	0.0083	0.9445
0.4801	50.0	100	0.4696	0.3183	0.4981	0.9492	nan	0.0062	0.9901	0.0	0.0057	0.9492
0.4744	55.0	110	0.4317	0.3187	0.4987	0.9543	nan	0.0018	0.9956	0.0	0.0017	0.9543
0.4494	60.0	120	0.3991	0.3189	0.4991	0.9555	nan	0.0013	0.9969	0.0	0.0012	0.9555
0.386	65.0	130	0.3737	0.3189	0.4990	0.9565	nan	0.0000	0.9980	0.0	0.0000	0.9565
0.3674	70.0	140	0.3538	0.3191	0.4994	0.9567	nan	0.0007	0.9981	0.0	0.0007	0.9567
0.3601	75.0	150	0.3413	0.3192	0.4995	0.9573	nan	0.0002	0.9988	0.0	0.0002	0.9573
0.3626	80.0	160	0.3225	0.3193	0.4996	0.9569	nan	0.0009	0.9984	0.0	0.0009	0.9569
0.3331	85.0	170	0.3163	0.3195	0.5000	0.9576	nan	0.0009	0.9991	0.0	0.0009	0.9576
0.319	90.0	180	0.3004	0.3200	0.5008	0.9577	nan	0.0024	0.9991	0.0	0.0024	0.9577
0.3163	95.0	190	0.2931	0.3198	0.5004	0.9575	nan	0.0020	0.9989	0.0	0.0020	0.9575
0.3185	100.0	200	0.2920	0.3194	0.4999	0.9577	nan	0.0006	0.9992	0.0	0.0006	0.9577
0.3122	105.0	210	0.2831	0.3194	0.4999	0.9578	nan	0.0005	0.9994	0.0	0.0005	0.9578
0.3218	110.0	220	0.2788	0.3195	0.5000	0.9576	nan	0.0009	0.9991	0.0	0.0009	0.9576
0.3037	115.0	230	0.2752	0.3194	0.4999	0.9577	nan	0.0006	0.9992	0.0	0.0006	0.9577
0.3319	120.0	240	0.2768	0.3194	0.4999	0.9578	nan	0.0006	0.9993	0.0	0.0006	0.9578

Framework versions

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