dropoff-utcustom-train-SF-RGB-b5_5

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: 0.1911
• Mean Iou: 0.4677
• Mean Accuracy: 0.7472
• Overall Accuracy: 0.9719
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.5020
• Accuracy Undropoff: 0.9923
• Iou Unlabeled: 0.0
• Iou Dropoff: 0.4318
• Iou Undropoff: 0.9713

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: 9e-06
• 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
1.0685	5.0	10	1.0222	0.2189	0.3725	0.5989	nan	0.1256	0.6194	0.0	0.0497	0.6070
0.9481	10.0	20	0.8419	0.3703	0.6398	0.8451	nan	0.4159	0.8637	0.0	0.2633	0.8476
0.8268	15.0	30	0.7165	0.3949	0.6938	0.8694	nan	0.5023	0.8853	0.0	0.3136	0.8711
0.7573	20.0	40	0.6206	0.4084	0.7186	0.8994	nan	0.5214	0.9158	0.0	0.3243	0.9010
0.636	25.0	50	0.5194	0.4239	0.7253	0.9300	nan	0.5020	0.9485	0.0	0.3401	0.9316
0.5238	30.0	60	0.4507	0.4365	0.7368	0.9461	nan	0.5085	0.9651	0.0	0.3618	0.9476
0.4296	35.0	70	0.4064	0.4410	0.7422	0.9530	nan	0.5123	0.9721	0.0	0.3683	0.9546
0.4105	40.0	80	0.3547	0.4502	0.7467	0.9619	nan	0.5120	0.9814	0.0	0.3880	0.9627
0.3436	45.0	90	0.3304	0.4571	0.7596	0.9644	nan	0.5361	0.9830	0.0	0.4066	0.9647
0.2729	50.0	100	0.2953	0.4614	0.7552	0.9680	nan	0.5232	0.9873	0.0	0.4163	0.9678
0.2546	55.0	110	0.2770	0.4629	0.7579	0.9691	nan	0.5276	0.9882	0.0	0.4201	0.9686
0.2281	60.0	120	0.2591	0.4647	0.7566	0.9702	nan	0.5235	0.9896	0.0	0.4245	0.9696
0.2041	65.0	130	0.2453	0.4657	0.7556	0.9708	nan	0.5209	0.9903	0.0	0.4269	0.9701
0.1772	70.0	140	0.2292	0.4676	0.7542	0.9717	nan	0.5171	0.9914	0.0	0.4317	0.9711
0.169	75.0	150	0.2161	0.4681	0.7520	0.9719	nan	0.5122	0.9919	0.0	0.4331	0.9713
0.1543	80.0	160	0.2111	0.4682	0.7530	0.9715	nan	0.5147	0.9913	0.0	0.4336	0.9709
0.1374	85.0	170	0.1973	0.4659	0.7450	0.9715	nan	0.4980	0.9921	0.0	0.4268	0.9709
0.1523	90.0	180	0.1974	0.4681	0.7501	0.9717	nan	0.5085	0.9918	0.0	0.4332	0.9711
0.1323	95.0	190	0.1928	0.4658	0.7434	0.9717	nan	0.4944	0.9924	0.0	0.4263	0.9711
0.1254	100.0	200	0.1923	0.4671	0.7467	0.9717	nan	0.5013	0.9921	0.0	0.4301	0.9711
0.125	105.0	210	0.1867	0.4637	0.7380	0.9717	nan	0.4831	0.9929	0.0	0.4201	0.9711
0.1239	110.0	220	0.1912	0.4694	0.7520	0.9719	nan	0.5121	0.9919	0.0	0.4369	0.9713
0.1252	115.0	230	0.1913	0.4689	0.7503	0.9720	nan	0.5085	0.9921	0.0	0.4354	0.9714
0.1357	120.0	240	0.1911	0.4677	0.7472	0.9719	nan	0.5020	0.9923	0.0	0.4318	0.9713

Framework versions

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