dropoff-utcustom-train-SF-RGBD-b0_4

This model is a fine-tuned version of nvidia/mit-b0 on the sam1120/dropoff-utcustom-TRAIN dataset. It achieves the following results on the evaluation set:

• Loss: 0.3688
• Mean Iou: 0.3485
• Mean Accuracy: 0.5433
• Overall Accuracy: 0.9606
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.0881
• Accuracy Undropoff: 0.9984
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.0851
• Iou Undropoff: 0.9604

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-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
1.2008	5.0	10	1.0960	0.1205	0.4461	0.2825	nan	0.6246	0.2677	0.0	0.0943	0.2671
1.0485	10.0	20	1.0952	0.1603	0.6272	0.4049	nan	0.8696	0.3848	0.0	0.0965	0.3843
0.9156	15.0	30	1.0312	0.3080	0.5963	0.8333	nan	0.3377	0.8548	0.0	0.0924	0.8317
0.7435	20.0	40	0.9448	0.3221	0.5508	0.8937	nan	0.1769	0.9248	0.0	0.0733	0.8930
0.7336	25.0	50	0.7446	0.3191	0.4998	0.9461	nan	0.0129	0.9866	0.0	0.0113	0.9461
0.6585	30.0	60	0.6397	0.3183	0.4981	0.9534	nan	0.0014	0.9948	0.0	0.0013	0.9534
0.583	35.0	70	0.5785	0.3181	0.4978	0.9537	nan	0.0006	0.9951	0.0	0.0005	0.9537
0.5324	40.0	80	0.5458	0.3182	0.4980	0.9545	nan	0.0002	0.9958	0.0	0.0002	0.9545
0.5155	45.0	90	0.5347	0.3186	0.4987	0.9558	nan	0.0001	0.9973	0.0	0.0001	0.9558
0.4874	50.0	100	0.4954	0.3179	0.4976	0.9537	nan	0.0	0.9951	0.0	0.0	0.9537
0.4716	55.0	110	0.4646	0.3185	0.4985	0.9555	nan	0.0	0.9969	0.0	0.0	0.9555
0.4441	60.0	120	0.4426	0.3185	0.4985	0.9555	nan	0.0	0.9970	0.0	0.0	0.9555
0.4659	65.0	130	0.4345	0.3189	0.4991	0.9567	nan	0.0	0.9982	0.0	0.0	0.9567
0.4758	70.0	140	0.4221	0.3181	0.4978	0.9543	nan	0.0	0.9957	0.0	0.0	0.9543
0.4208	75.0	150	0.4029	0.3190	0.4993	0.9571	nan	0.0	0.9987	0.0	0.0	0.9571
0.4395	80.0	160	0.4170	0.3207	0.5016	0.9559	nan	0.0062	0.9971	0.0	0.0062	0.9559
0.3981	85.0	170	0.3992	0.3214	0.5027	0.9574	nan	0.0067	0.9987	0.0	0.0066	0.9574
0.3983	90.0	180	0.3965	0.3282	0.5125	0.9560	nan	0.0288	0.9963	0.0	0.0285	0.9560
0.398	95.0	190	0.3747	0.3272	0.5112	0.9569	nan	0.0251	0.9973	0.0	0.0249	0.9568
0.3767	100.0	200	0.3722	0.3301	0.5155	0.9574	nan	0.0336	0.9975	0.0	0.0330	0.9573
0.3797	105.0	210	0.3781	0.3334	0.5204	0.9583	nan	0.0429	0.9980	0.0	0.0420	0.9582
0.373	110.0	220	0.3744	0.3409	0.5317	0.9593	nan	0.0654	0.9980	0.0	0.0636	0.9591
0.372	115.0	230	0.3700	0.3440	0.5364	0.9599	nan	0.0746	0.9983	0.0	0.0723	0.9598
0.3629	120.0	240	0.3688	0.3485	0.5433	0.9606	nan	0.0881	0.9984	0.0	0.0851	0.9604

Framework versions

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