dropoff-utcustom-train-SF-RGBD-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: 0.1429
• Mean Iou: 0.6443
• Mean Accuracy: 0.6853
• Overall Accuracy: 0.9669
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.3782
• Accuracy Undropoff: 0.9925
• Iou Unlabeled: nan
• Iou Dropoff: 0.3223
• Iou Undropoff: 0.9664

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
1.159	5.0	10	1.0040	0.2283	0.5676	0.6267	nan	0.5031	0.6321	0.0	0.0644	0.6203
0.8345	10.0	20	0.7480	0.3236	0.5320	0.9158	nan	0.1134	0.9506	0.0	0.0555	0.9154
0.5406	15.0	30	0.5477	0.3223	0.5049	0.9513	nan	0.0179	0.9918	0.0	0.0157	0.9513
0.3695	20.0	40	0.4590	0.3215	0.5036	0.9519	nan	0.0146	0.9926	0.0	0.0125	0.9519
0.3053	25.0	50	0.3790	0.3196	0.5001	0.9565	nan	0.0023	0.9979	0.0	0.0022	0.9565
0.2436	30.0	60	0.3303	0.4812	0.5020	0.9568	nan	0.0059	0.9981	nan	0.0056	0.9568
0.2148	35.0	70	0.2739	0.4794	0.5002	0.9580	nan	0.0008	0.9996	nan	0.0008	0.9580
0.1983	40.0	80	0.2348	0.5079	0.5284	0.9595	nan	0.0582	0.9986	nan	0.0564	0.9594
0.1784	45.0	90	0.2178	0.6064	0.6440	0.9631	nan	0.2960	0.9920	nan	0.2501	0.9626
0.1631	50.0	100	0.1943	0.6223	0.6811	0.9607	nan	0.3760	0.9861	nan	0.2846	0.9601
0.1468	55.0	110	0.1759	0.6206	0.6731	0.9617	nan	0.3583	0.9879	nan	0.2801	0.9611
0.1353	60.0	120	0.1657	0.6014	0.6335	0.9639	nan	0.2731	0.9939	nan	0.2393	0.9635
0.1474	65.0	130	0.1590	0.5943	0.6228	0.9641	nan	0.2505	0.9951	nan	0.2249	0.9637
0.1172	70.0	140	0.1562	0.6272	0.6662	0.9653	nan	0.3400	0.9924	nan	0.2896	0.9648
0.1169	75.0	150	0.1538	0.6302	0.6696	0.9656	nan	0.3467	0.9925	nan	0.2954	0.9651
0.1263	80.0	160	0.1540	0.6372	0.6784	0.9661	nan	0.3645	0.9922	nan	0.3089	0.9656
0.1028	85.0	170	0.1512	0.6462	0.6948	0.9659	nan	0.3992	0.9904	nan	0.3271	0.9653
0.1163	90.0	180	0.1493	0.6469	0.6932	0.9663	nan	0.3953	0.9911	nan	0.3280	0.9658
0.0998	95.0	190	0.1481	0.6457	0.6894	0.9666	nan	0.3869	0.9918	nan	0.3253	0.9661
0.0997	100.0	200	0.1465	0.6454	0.6893	0.9665	nan	0.3869	0.9917	nan	0.3247	0.9660
0.0998	105.0	210	0.1473	0.6488	0.6937	0.9668	nan	0.3958	0.9916	nan	0.3313	0.9662
0.1003	110.0	220	0.1437	0.6401	0.6774	0.9671	nan	0.3614	0.9934	nan	0.3136	0.9666
0.0932	115.0	230	0.1434	0.6469	0.6898	0.9669	nan	0.3876	0.9920	nan	0.3275	0.9664
0.0942	120.0	240	0.1429	0.6443	0.6853	0.9669	nan	0.3782	0.9925	nan	0.3223	0.9664

Framework versions

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