dropoff-utcustom-train-SF-RGBD-b5_5

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: 0.2636
• Mean Iou: 0.4256
• Mean Accuracy: 0.6832
• Overall Accuracy: 0.9656
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.3752
• Accuracy Undropoff: 0.9912
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.3118
• Iou Undropoff: 0.9650

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: 9e-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
1.1487	5.0	10	1.0250	0.2562	0.6276	0.6778	nan	0.5730	0.6823	0.0	0.0971	0.6714
1.0128	10.0	20	0.9030	0.3142	0.6730	0.8268	nan	0.5053	0.8407	0.0	0.1195	0.8231
0.8561	15.0	30	0.7359	0.3520	0.6913	0.8949	nan	0.4692	0.9133	0.0	0.1632	0.8928
0.7551	20.0	40	0.6534	0.3634	0.6999	0.9090	nan	0.4719	0.9280	0.0	0.1829	0.9072
0.6236	25.0	50	0.5938	0.3710	0.7001	0.9189	nan	0.4614	0.9388	0.0	0.1955	0.9173
0.4977	30.0	60	0.5293	0.3850	0.6987	0.9341	nan	0.4420	0.9555	0.0	0.2222	0.9329
0.4188	35.0	70	0.4859	0.3935	0.6941	0.9425	nan	0.4231	0.9650	0.0	0.2390	0.9415
0.3532	40.0	80	0.4278	0.4019	0.6823	0.9519	nan	0.3881	0.9764	0.0	0.2547	0.9511
0.3187	45.0	90	0.3914	0.4098	0.6873	0.9560	nan	0.3942	0.9804	0.0	0.2742	0.9553
0.2631	50.0	100	0.3647	0.4134	0.6918	0.9575	nan	0.4020	0.9815	0.0	0.2835	0.9567
0.2565	55.0	110	0.3424	0.4141	0.6895	0.9585	nan	0.3962	0.9829	0.0	0.2846	0.9578
0.2259	60.0	120	0.3127	0.4178	0.6853	0.9613	nan	0.3843	0.9863	0.0	0.2926	0.9607
0.2263	65.0	130	0.2920	0.4202	0.6822	0.9632	nan	0.3757	0.9886	0.0	0.2981	0.9626
0.1961	70.0	140	0.2755	0.4218	0.6769	0.9649	nan	0.3627	0.9911	0.0	0.3009	0.9644
0.1897	75.0	150	0.2726	0.4232	0.6803	0.9650	nan	0.3698	0.9908	0.0	0.3052	0.9645
0.1863	80.0	160	0.2762	0.4241	0.6830	0.9649	nan	0.3756	0.9904	0.0	0.3079	0.9643
0.1656	85.0	170	0.2730	0.4241	0.6809	0.9653	nan	0.3708	0.9911	0.0	0.3076	0.9648
0.1745	90.0	180	0.2740	0.4241	0.6821	0.9651	nan	0.3736	0.9907	0.0	0.3079	0.9645
0.1726	95.0	190	0.2779	0.4242	0.6854	0.9645	nan	0.3809	0.9898	0.0	0.3085	0.9639
0.158	100.0	200	0.2661	0.4248	0.6808	0.9656	nan	0.3701	0.9915	0.0	0.3094	0.9651
0.19	105.0	210	0.2667	0.4240	0.6790	0.9656	nan	0.3664	0.9916	0.0	0.3070	0.9651
0.1533	110.0	220	0.2696	0.4258	0.6843	0.9655	nan	0.3777	0.9910	0.0	0.3126	0.9649
0.1644	115.0	230	0.2690	0.4261	0.6855	0.9654	nan	0.3803	0.9908	0.0	0.3136	0.9648
0.1594	120.0	240	0.2636	0.4256	0.6832	0.9656	nan	0.3752	0.9912	0.0	0.3118	0.9650

Framework versions

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