dropoff-utcustom-train-SF-RGB-b0_3

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: 0.3958
• Mean Iou: 0.6134
• Mean Accuracy: 0.6480
• Overall Accuracy: 0.9627
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.3026
• Accuracy Undropoff: 0.9933
• Iou Unlabeled: nan
• Iou Dropoff: 0.2645
• Iou Undropoff: 0.9622

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-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
1.1015	3.33	10	1.0990	0.1184	0.4572	0.3294	nan	0.5975	0.3170	0.0	0.0427	0.3124
1.0478	6.67	20	1.0756	0.2121	0.7082	0.5654	nan	0.8648	0.5515	0.0	0.0879	0.5482
1.0451	10.0	30	1.0269	0.2846	0.8053	0.7334	nan	0.8842	0.7264	0.0	0.1313	0.7226
0.9095	13.33	40	0.9476	0.3360	0.7905	0.8411	nan	0.7349	0.8460	0.0	0.1723	0.8358
0.8091	16.67	50	0.8425	0.3858	0.7645	0.9167	nan	0.5975	0.9315	0.0	0.2429	0.9145
0.8094	20.0	60	0.7489	0.4090	0.7445	0.9417	nan	0.5281	0.9608	0.0	0.2866	0.9403
0.6945	23.33	70	0.7005	0.4148	0.7472	0.9453	nan	0.5298	0.9646	0.0	0.3004	0.9440
0.6337	26.67	80	0.6331	0.6267	0.7334	0.9499	nan	0.4958	0.9709	nan	0.3047	0.9488
0.603	30.0	90	0.5726	0.6222	0.6935	0.9559	nan	0.4057	0.9814	nan	0.2894	0.9551
0.5903	33.33	100	0.5841	0.6248	0.7151	0.9526	nan	0.4546	0.9757	nan	0.2980	0.9516
0.5514	36.67	110	0.5157	0.6227	0.6818	0.9585	nan	0.3781	0.9854	nan	0.2875	0.9578
0.6464	40.0	120	0.5141	0.6240	0.6889	0.9575	nan	0.3941	0.9836	nan	0.2912	0.9568
0.5198	43.33	130	0.4890	0.4141	0.6762	0.9591	nan	0.3657	0.9866	0.0	0.2838	0.9585
0.5077	46.67	140	0.4855	0.4118	0.6719	0.9588	nan	0.3572	0.9866	0.0	0.2773	0.9581
0.4817	50.0	150	0.4710	0.6182	0.6733	0.9587	nan	0.3602	0.9864	nan	0.2784	0.9580
0.4713	53.33	160	0.4669	0.6196	0.6683	0.9603	nan	0.3479	0.9887	nan	0.2795	0.9597
0.4516	56.67	170	0.4486	0.4107	0.6586	0.9612	nan	0.3265	0.9906	0.0	0.2715	0.9606
0.4059	60.0	180	0.4361	0.6136	0.6548	0.9612	nan	0.3187	0.9909	nan	0.2665	0.9606
0.4142	63.33	190	0.4267	0.6115	0.6503	0.9615	nan	0.3089	0.9917	nan	0.2621	0.9610
0.4393	66.67	200	0.4188	0.6035	0.6354	0.9623	nan	0.2768	0.9940	nan	0.2452	0.9618
0.4071	70.0	210	0.4224	0.6137	0.6528	0.9617	nan	0.3138	0.9917	nan	0.2663	0.9612
0.4009	73.33	220	0.4205	0.6136	0.6540	0.9614	nan	0.3167	0.9912	nan	0.2664	0.9608
0.4043	76.67	230	0.4148	0.6132	0.6514	0.9619	nan	0.3108	0.9920	nan	0.2651	0.9613
0.6302	80.0	240	0.4116	0.6133	0.6513	0.9619	nan	0.3105	0.9921	nan	0.2653	0.9614
0.3859	83.33	250	0.4113	0.6141	0.6543	0.9615	nan	0.3174	0.9913	nan	0.2673	0.9609
0.3791	86.67	260	0.4033	0.6042	0.6361	0.9623	nan	0.2782	0.9940	nan	0.2465	0.9619
0.5716	90.0	270	0.4088	0.6168	0.6575	0.9617	nan	0.3237	0.9913	nan	0.2724	0.9612
0.3803	93.33	280	0.4024	0.6171	0.6565	0.9621	nan	0.3211	0.9918	nan	0.2727	0.9615
0.371	96.67	290	0.3979	0.6166	0.6539	0.9625	nan	0.3154	0.9925	nan	0.2713	0.9620
0.3656	100.0	300	0.3992	0.6204	0.6615	0.9621	nan	0.3316	0.9913	nan	0.2793	0.9615
0.3674	103.33	310	0.3930	0.6110	0.6433	0.9630	nan	0.2925	0.9941	nan	0.2594	0.9625
0.378	106.67	320	0.3925	0.6124	0.6459	0.9629	nan	0.2981	0.9937	nan	0.2623	0.9624
0.5766	110.0	330	0.3965	0.6192	0.6594	0.9621	nan	0.3272	0.9916	nan	0.2768	0.9616
0.3513	113.33	340	0.3927	0.6161	0.6523	0.9627	nan	0.3118	0.9928	nan	0.2701	0.9622
0.3731	116.67	350	0.3975	0.6200	0.6613	0.9620	nan	0.3315	0.9912	nan	0.2785	0.9614
0.3489	120.0	360	0.3958	0.6134	0.6480	0.9627	nan	0.3026	0.9933	nan	0.2645	0.9622

Framework versions

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