dropoff-utcustom-train-SF-RGB-b5_3

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: 0.3770
• Mean Iou: 0.4572
• Mean Accuracy: 0.7822
• Overall Accuracy: 0.9640
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.5839
• Accuracy Undropoff: 0.9805
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.4086
• Iou Undropoff: 0.9631

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-06
• train_batch_size: 15
• eval_batch_size: 15
• 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
1.3135	5.0	10	1.2008	0.0546	0.2586	0.1227	nan	0.4069	0.1103	0.0	0.0535	0.1102
1.2309	10.0	20	1.1294	0.1176	0.3397	0.2490	nan	0.4388	0.2407	0.0	0.1129	0.2400
1.1346	15.0	30	1.0395	0.2171	0.4865	0.5022	nan	0.4694	0.5036	0.0	0.1524	0.4989
1.1088	20.0	40	0.9755	0.2608	0.5521	0.6176	nan	0.4808	0.6235	0.0	0.1661	0.6163
1.007	25.0	50	0.9197	0.2895	0.5959	0.6775	nan	0.5068	0.6849	0.0	0.1923	0.6763
0.9145	30.0	60	0.8635	0.3162	0.6299	0.7335	nan	0.5168	0.7429	0.0	0.2156	0.7329
0.8745	35.0	70	0.8070	0.3398	0.6784	0.7808	nan	0.5667	0.7901	0.0	0.2404	0.7791
0.8088	40.0	80	0.7442	0.3667	0.7191	0.8290	nan	0.5993	0.8389	0.0	0.2730	0.8272
0.7184	45.0	90	0.6956	0.3832	0.7513	0.8603	nan	0.6323	0.8702	0.0	0.2915	0.8580
0.6908	50.0	100	0.6751	0.3931	0.7592	0.8748	nan	0.6332	0.8853	0.0	0.3067	0.8728
0.643	55.0	110	0.6101	0.4134	0.7714	0.9108	nan	0.6194	0.9234	0.0	0.3308	0.9094
0.6014	60.0	120	0.5971	0.4166	0.7826	0.9189	nan	0.6339	0.9313	0.0	0.3324	0.9175
0.5685	65.0	130	0.5595	0.4304	0.7946	0.9328	nan	0.6439	0.9453	0.0	0.3599	0.9314
0.5172	70.0	140	0.5344	0.4373	0.8010	0.9406	nan	0.6488	0.9532	0.0	0.3727	0.9393
0.4757	75.0	150	0.4963	0.4434	0.7997	0.9490	nan	0.6368	0.9626	0.0	0.3822	0.9479
0.4288	80.0	160	0.4599	0.4488	0.7936	0.9556	nan	0.6169	0.9702	0.0	0.3918	0.9546
0.4124	85.0	170	0.4710	0.4469	0.7989	0.9540	nan	0.6296	0.9681	0.0	0.3876	0.9529
0.4995	90.0	180	0.4209	0.4537	0.7883	0.9606	nan	0.6004	0.9762	0.0	0.4015	0.9597
0.3815	95.0	190	0.4287	0.4524	0.7919	0.9595	nan	0.6090	0.9748	0.0	0.3988	0.9586
0.3764	100.0	200	0.4245	0.4529	0.7913	0.9600	nan	0.6073	0.9753	0.0	0.3998	0.9590
0.4074	105.0	210	0.4096	0.4542	0.7894	0.9613	nan	0.6018	0.9769	0.0	0.4021	0.9603
0.3975	110.0	220	0.4107	0.4538	0.7905	0.9610	nan	0.6045	0.9765	0.0	0.4013	0.9601
0.3598	115.0	230	0.3918	0.4558	0.7863	0.9627	nan	0.5939	0.9787	0.0	0.4057	0.9618
0.3709	120.0	240	0.3770	0.4572	0.7822	0.9640	nan	0.5839	0.9805	0.0	0.4086	0.9631

Framework versions

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