dropoff-utcustom-train-SF-RGB-b5_1

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: 0.6279
• Mean Iou: 0.4054
• Mean Accuracy: 0.7471
• Overall Accuracy: 0.8860
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.5956
• Accuracy Undropoff: 0.8986
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.3318
• Iou Undropoff: 0.8843

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-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
1.0071	5.0	10	1.0206	0.1745	0.2748	0.5034	nan	0.0255	0.5241	0.0	0.0147	0.5087
0.9688	10.0	20	0.9873	0.2140	0.3486	0.5771	nan	0.0992	0.5979	0.0	0.0582	0.5838
0.9406	15.0	30	0.9313	0.2613	0.4446	0.6655	nan	0.2038	0.6855	0.0	0.1135	0.6705
0.9278	20.0	40	0.8851	0.2930	0.5149	0.7111	nan	0.3009	0.7289	0.0	0.1648	0.7142
0.8956	25.0	50	0.8563	0.3118	0.5642	0.7358	nan	0.3770	0.7514	0.0	0.1985	0.7370
0.8674	30.0	60	0.8260	0.3303	0.6086	0.7664	nan	0.4366	0.7807	0.0	0.2246	0.7664
0.8438	35.0	70	0.8149	0.3347	0.6355	0.7671	nan	0.4921	0.7790	0.0	0.2381	0.7660
0.8309	40.0	80	0.7881	0.3459	0.6472	0.7847	nan	0.4972	0.7972	0.0	0.2539	0.7839
0.8069	45.0	90	0.7640	0.3567	0.6617	0.8041	nan	0.5063	0.8170	0.0	0.2668	0.8033
0.7779	50.0	100	0.7486	0.3637	0.6792	0.8145	nan	0.5316	0.8268	0.0	0.2778	0.8132
0.7695	55.0	110	0.7354	0.3684	0.6936	0.8214	nan	0.5542	0.8329	0.0	0.2858	0.8195
0.7568	60.0	120	0.7164	0.3757	0.7032	0.8365	nan	0.5577	0.8486	0.0	0.2924	0.8347
0.7285	65.0	130	0.6976	0.3836	0.7119	0.8484	nan	0.5630	0.8608	0.0	0.3042	0.8467
0.7217	70.0	140	0.6922	0.3857	0.7217	0.8499	nan	0.5817	0.8616	0.0	0.3091	0.8480
0.7095	75.0	150	0.6708	0.3926	0.7287	0.8624	nan	0.5828	0.8745	0.0	0.3172	0.8605
0.6944	80.0	160	0.6637	0.3951	0.7320	0.8660	nan	0.5858	0.8781	0.0	0.3212	0.8641
0.6878	85.0	170	0.6632	0.3942	0.7397	0.8673	nan	0.6005	0.8788	0.0	0.3175	0.8652
0.6868	90.0	180	0.6468	0.3998	0.7391	0.8756	nan	0.5902	0.8880	0.0	0.3257	0.8739
0.6581	95.0	190	0.6444	0.4003	0.7421	0.8776	nan	0.5942	0.8899	0.0	0.3249	0.8759
0.6587	100.0	200	0.6383	0.4026	0.7427	0.8814	nan	0.5914	0.8940	0.0	0.3281	0.8797
0.6525	105.0	210	0.6334	0.4032	0.7434	0.8825	nan	0.5918	0.8951	0.0	0.3289	0.8808
0.658	110.0	220	0.6345	0.4026	0.7451	0.8811	nan	0.5968	0.8934	0.0	0.3285	0.8793
0.6575	115.0	230	0.6300	0.4050	0.7463	0.8851	nan	0.5948	0.8977	0.0	0.3314	0.8835
0.6625	120.0	240	0.6279	0.4054	0.7471	0.8860	nan	0.5956	0.8986	0.0	0.3318	0.8843

Framework versions

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