dropoff-utcustom-train-SF-RGB-b0_2

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

• Loss: 0.6222
• Mean Iou: 0.4086
• Mean Accuracy: 0.6638
• Overall Accuracy: 0.9583
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.3408
• Accuracy Undropoff: 0.9869
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.2682
• Iou Undropoff: 0.9576

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: 1e-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
1.1857	3.33	10	1.1215	0.0852	0.2039	0.0995	nan	0.3183	0.0894	0.0	0.1663	0.0893
1.1597	6.67	20	1.1165	0.1437	0.3568	0.2326	nan	0.4930	0.2206	0.0	0.2108	0.2203
1.1528	10.0	30	1.1040	0.1938	0.5140	0.3464	nan	0.6978	0.3301	0.0	0.2517	0.3297
1.0852	13.33	40	1.0896	0.2243	0.6289	0.4219	nan	0.8561	0.4018	0.0	0.2717	0.4011
1.0388	16.67	50	1.0511	0.2700	0.6748	0.5498	nan	0.8120	0.5377	0.0	0.2744	0.5357
1.0426	20.0	60	1.0089	0.3147	0.6787	0.6640	nan	0.6949	0.6625	0.0	0.2857	0.6583
0.9621	23.33	70	0.9921	0.3374	0.7060	0.7392	nan	0.6695	0.7424	0.0	0.2760	0.7361
0.925	26.67	80	0.9464	0.3591	0.7031	0.7964	nan	0.6007	0.8054	0.0	0.2807	0.7965
0.8872	30.0	90	0.8993	0.3858	0.7074	0.8676	nan	0.5316	0.8831	0.0	0.2888	0.8686
0.8751	33.33	100	0.8974	0.3896	0.7177	0.8817	nan	0.5379	0.8976	0.0	0.2866	0.8822
0.8571	36.67	110	0.8501	0.4028	0.7162	0.9122	nan	0.5011	0.9312	0.0	0.2953	0.9131
0.8866	40.0	120	0.8434	0.4072	0.7240	0.9252	nan	0.5032	0.9448	0.0	0.2963	0.9254
0.8127	43.33	130	0.7922	0.4142	0.7089	0.9404	nan	0.4548	0.9629	0.0	0.3025	0.9402
0.8062	46.67	140	0.7917	0.4123	0.7103	0.9432	nan	0.4548	0.9658	0.0	0.2943	0.9425
0.7512	50.0	150	0.7646	0.4142	0.7059	0.9478	nan	0.4404	0.9713	0.0	0.2955	0.9470
0.7554	53.33	160	0.7497	0.4161	0.7001	0.9510	nan	0.4248	0.9754	0.0	0.2981	0.9502
0.7468	56.67	170	0.7326	0.4177	0.6989	0.9535	nan	0.4195	0.9782	0.0	0.3005	0.9527
0.6506	60.0	180	0.7184	0.4173	0.6992	0.9541	nan	0.4196	0.9789	0.0	0.2987	0.9533
0.6761	63.33	190	0.7037	0.4142	0.6884	0.9546	nan	0.3964	0.9805	0.0	0.2886	0.9539
0.7245	66.67	200	0.6960	0.4122	0.6821	0.9553	nan	0.3824	0.9818	0.0	0.2820	0.9545
0.6514	70.0	210	0.6755	0.4104	0.6705	0.9573	nan	0.3559	0.9852	0.0	0.2746	0.9566
0.6433	73.33	220	0.6804	0.4180	0.6954	0.9556	nan	0.4100	0.9809	0.0	0.2991	0.9548
0.6686	76.67	230	0.6608	0.4107	0.6694	0.9578	nan	0.3531	0.9858	0.0	0.2749	0.9571
0.9091	80.0	240	0.6701	0.4160	0.6922	0.9557	nan	0.4031	0.9813	0.0	0.2930	0.9549
0.6346	83.33	250	0.6725	0.4166	0.6904	0.9563	nan	0.3987	0.9821	0.0	0.2944	0.9555
0.6303	86.67	260	0.6460	0.4090	0.6670	0.9576	nan	0.3481	0.9858	0.0	0.2702	0.9569
0.8923	90.0	270	0.6550	0.4131	0.6799	0.9568	nan	0.3760	0.9837	0.0	0.2832	0.9561
0.6334	93.33	280	0.6468	0.4100	0.6708	0.9572	nan	0.3566	0.9851	0.0	0.2734	0.9565
0.6242	96.67	290	0.6483	0.4106	0.6728	0.9572	nan	0.3607	0.9848	0.0	0.2754	0.9565
0.7401	100.0	300	0.6470	0.4129	0.6796	0.9569	nan	0.3755	0.9838	0.0	0.2825	0.9561
0.6148	103.33	310	0.6242	0.4081	0.6633	0.9582	nan	0.3397	0.9868	0.0	0.2668	0.9575
0.6345	106.67	320	0.6287	0.4093	0.6670	0.9579	nan	0.3478	0.9862	0.0	0.2708	0.9573
0.8711	110.0	330	0.6396	0.4130	0.6782	0.9572	nan	0.3720	0.9843	0.0	0.2826	0.9565
0.5812	113.33	340	0.6266	0.4101	0.6689	0.9580	nan	0.3517	0.9861	0.0	0.2731	0.9573
0.6503	116.67	350	0.6384	0.4130	0.6775	0.9573	nan	0.3706	0.9845	0.0	0.2824	0.9566
0.5923	120.0	360	0.6222	0.4086	0.6638	0.9583	nan	0.3408	0.9869	0.0	0.2682	0.9576

Framework versions

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