safety-utcustom-train-SF30-RGBD-b0

This model is a fine-tuned version of nvidia/mit-b0 on the sam1120/safety-utcustom-TRAIN-30 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3227
• Mean Iou: 0.5786
• Mean Accuracy: 0.6222
• Overall Accuracy: 0.9658
• Accuracy Unlabeled: nan
• Accuracy Safe: 0.2552
• Accuracy Unsafe: 0.9891
• Iou Unlabeled: nan
• Iou Safe: 0.1917
• Iou Unsafe: 0.9655

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: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Safe	Accuracy Unsafe	Iou Unlabeled	Iou Safe	Iou Unsafe
0.9925	5.0	10	1.0612	0.3101	0.5355	0.8847	nan	0.1625	0.9085	0.0	0.0462	0.8841
0.8589	10.0	20	0.9441	0.3303	0.5181	0.9537	nan	0.0529	0.9833	0.0	0.0373	0.9537
0.7016	15.0	30	0.7764	0.3274	0.5069	0.9654	nan	0.0172	0.9965	0.0	0.0169	0.9654
0.6093	20.0	40	0.6213	0.3339	0.5219	0.9603	nan	0.0538	0.9901	0.0	0.0415	0.9603
0.5281	25.0	50	0.5431	0.3355	0.5213	0.9650	nan	0.0476	0.9951	0.0	0.0417	0.9649
0.5077	30.0	60	0.5043	0.3361	0.5231	0.9638	nan	0.0524	0.9938	0.0	0.0444	0.9638
0.5197	35.0	70	0.4579	0.3379	0.5249	0.9657	nan	0.0543	0.9956	0.0	0.0481	0.9656
0.4477	40.0	80	0.4340	0.3395	0.5271	0.9662	nan	0.0583	0.9960	0.0	0.0523	0.9661
0.4371	45.0	90	0.4033	0.3407	0.5287	0.9669	nan	0.0607	0.9967	0.0	0.0553	0.9669
0.3972	50.0	100	0.3975	0.3420	0.5292	0.9686	nan	0.0600	0.9985	0.0	0.0574	0.9686
0.4101	55.0	110	0.3777	0.5215	0.5381	0.9691	nan	0.0778	0.9983	nan	0.0740	0.9690
0.3528	60.0	120	0.3625	0.5360	0.5587	0.9668	nan	0.1229	0.9945	nan	0.1054	0.9667
0.3552	65.0	130	0.3733	0.5550	0.5829	0.9671	nan	0.1726	0.9932	nan	0.1430	0.9669
0.3798	70.0	140	0.3444	0.5598	0.5753	0.9722	nan	0.1515	0.9991	nan	0.1476	0.9720
0.3235	75.0	150	0.3461	0.5651	0.6041	0.9650	nan	0.2187	0.9895	nan	0.1656	0.9647
0.3457	80.0	160	0.3335	0.5638	0.5880	0.9695	nan	0.1806	0.9954	nan	0.1582	0.9693
0.318	85.0	170	0.3334	0.5739	0.6114	0.9667	nan	0.2321	0.9908	nan	0.1814	0.9665
0.32	90.0	180	0.3307	0.5779	0.6112	0.9684	nan	0.2299	0.9926	nan	0.1877	0.9681
0.3122	95.0	190	0.3263	0.5778	0.6175	0.9667	nan	0.2447	0.9904	nan	0.1891	0.9664
0.3554	100.0	200	0.3227	0.5786	0.6222	0.9658	nan	0.2552	0.9891	nan	0.1917	0.9655

Framework versions

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