README.md

metadata

license: other
tags:
  - vision
  - image-segmentation
  - generated_from_trainer
model-index:
  - name: safety-utcustom-train-SF-RGBD-b0
    results: []

safety-utcustom-train-SF-RGBD-b0

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

• Loss: 0.2207
• Mean Iou: 0.6197
• Mean Accuracy: 0.6401
• Overall Accuracy: 0.9766
• Accuracy Unlabeled: nan
• Accuracy Safe: 0.2824
• Accuracy Unsafe: 0.9978
• Iou Unlabeled: nan
• Iou Safe: 0.2631
• Iou Unsafe: 0.9764

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: 30

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
1.0084	1.0	10	1.0688	0.2610	0.4107	0.7625	nan	0.0368	0.7845	0.0	0.0163	0.7666
0.8483	2.0	20	0.8740	0.3230	0.4991	0.9686	nan	0.0002	0.9980	0.0	0.0002	0.9687
0.7058	3.0	30	0.7416	0.3217	0.4969	0.9637	nan	0.0009	0.9930	0.0	0.0009	0.9641
0.578	4.0	40	0.5969	0.3223	0.4980	0.9659	nan	0.0007	0.9953	0.0	0.0007	0.9662
0.5531	5.0	50	0.5068	0.3247	0.5018	0.9681	nan	0.0061	0.9974	0.0	0.0059	0.9682
0.4786	6.0	60	0.4575	0.3254	0.5029	0.9670	nan	0.0097	0.9961	0.0	0.0092	0.9671
0.4681	7.0	70	0.4382	0.3251	0.5025	0.9690	nan	0.0067	0.9983	0.0	0.0064	0.9690
0.4139	8.0	80	0.3973	0.3234	0.4998	0.9686	nan	0.0017	0.9980	0.0	0.0016	0.9686
0.4275	9.0	90	0.3983	0.4888	0.5036	0.9701	nan	0.0077	0.9994	nan	0.0076	0.9701
0.3975	10.0	100	0.3398	0.3237	0.5003	0.9702	nan	0.0008	0.9998	0.0	0.0008	0.9702
0.4325	11.0	110	0.3785	0.3548	0.5467	0.9725	nan	0.0941	0.9993	0.0	0.0919	0.9725
0.3239	12.0	120	0.3338	0.3493	0.5383	0.9722	nan	0.0772	0.9995	0.0	0.0759	0.9722
0.3733	13.0	130	0.3013	0.5236	0.5379	0.9722	nan	0.0763	0.9995	nan	0.0751	0.9722
0.3165	14.0	140	0.2849	0.5254	0.5397	0.9723	nan	0.0800	0.9994	nan	0.0786	0.9722
0.3329	15.0	150	0.3002	0.5405	0.5554	0.9728	nan	0.1118	0.9990	nan	0.1083	0.9727
0.3214	16.0	160	0.2725	0.5309	0.5451	0.9726	nan	0.0908	0.9995	nan	0.0892	0.9726
0.2744	17.0	170	0.2896	0.5620	0.5780	0.9737	nan	0.1573	0.9986	nan	0.1503	0.9736
0.2948	18.0	180	0.2564	0.5507	0.5659	0.9733	nan	0.1330	0.9989	nan	0.1282	0.9732
0.2653	19.0	190	0.2518	0.5701	0.5860	0.9743	nan	0.1732	0.9987	nan	0.1660	0.9742
0.3026	20.0	200	0.2531	0.5550	0.5699	0.9737	nan	0.1408	0.9990	nan	0.1364	0.9735
0.2649	21.0	210	0.2384	0.5732	0.5894	0.9744	nan	0.1802	0.9986	nan	0.1722	0.9743
0.2431	22.0	220	0.2390	0.5818	0.5988	0.9747	nan	0.1993	0.9983	nan	0.1890	0.9746
0.2608	23.0	230	0.2355	0.5967	0.6149	0.9755	nan	0.2317	0.9981	nan	0.2181	0.9753
0.223	24.0	240	0.2290	0.5690	0.5843	0.9744	nan	0.1697	0.9989	nan	0.1637	0.9743
0.2448	25.0	250	0.2262	0.5894	0.6063	0.9753	nan	0.2141	0.9985	nan	0.2037	0.9751
0.2547	26.0	260	0.2281	0.6159	0.6357	0.9764	nan	0.2737	0.9978	nan	0.2555	0.9763
0.2266	27.0	270	0.2191	0.6004	0.6186	0.9757	nan	0.2391	0.9981	nan	0.2252	0.9755
0.2357	28.0	280	0.2218	0.5938	0.6106	0.9756	nan	0.2227	0.9985	nan	0.2122	0.9754
0.2239	29.0	290	0.2199	0.6138	0.6332	0.9764	nan	0.2686	0.9979	nan	0.2514	0.9762
0.2311	30.0	300	0.2207	0.6197	0.6401	0.9766	nan	0.2824	0.9978	nan	0.2631	0.9764

Framework versions

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