segformer_b1_finetuned_segment_pv_p100_16batch

This model is a fine-tuned version of nvidia/segformer-b1-finetuned-ade-512-512 on the mouadenna/satellite_PV_dataset_train_test_v1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0062
• Mean Iou: 0.8656
• Precision: 0.9155

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: 0.00016
• 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: 0.1
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Precision
0.59	1.0	230	0.2289	0.5149	0.5478
0.111	2.0	460	0.0320	0.7322	0.8038
0.0254	3.0	690	0.0133	0.7865	0.8738
0.0115	4.0	920	0.0079	0.8335	0.8829
0.0078	5.0	1150	0.0076	0.8156	0.8598
0.0061	6.0	1380	0.0061	0.8436	0.8926
0.0051	7.0	1610	0.0056	0.8478	0.9170
0.0042	8.0	1840	0.0059	0.8497	0.8975
0.0038	9.0	2070	0.0062	0.8431	0.9186
0.0037	10.0	2300	0.0055	0.8529	0.9142
0.0036	11.0	2530	0.0061	0.8397	0.8834
0.0035	12.0	2760	0.0055	0.8497	0.8981
0.0032	13.0	2990	0.0055	0.8485	0.9015
0.0028	14.0	3220	0.0056	0.8549	0.8979
0.0028	15.0	3450	0.0059	0.8523	0.8975
0.0026	16.0	3680	0.0055	0.8579	0.9120
0.0026	17.0	3910	0.0056	0.8587	0.9110
0.0024	18.0	4140	0.0074	0.8295	0.9233
0.0029	19.0	4370	0.0058	0.8548	0.9092
0.0025	20.0	4600	0.0055	0.8556	0.8914
0.0025	21.0	4830	0.0054	0.8569	0.9017
0.0028	22.0	5060	0.0055	0.8622	0.9166
0.0024	23.0	5290	0.0057	0.8633	0.9216
0.0022	24.0	5520	0.0059	0.8623	0.9155
0.002	25.0	5750	0.0060	0.8614	0.9046
0.002	26.0	5980	0.0062	0.8563	0.9092
0.0019	27.0	6210	0.0059	0.8642	0.9125
0.0018	28.0	6440	0.0060	0.8656	0.9097
0.0018	29.0	6670	0.0060	0.8632	0.9174
0.0018	30.0	6900	0.0061	0.8647	0.9172
0.0018	31.0	7130	0.0062	0.8657	0.9155
0.0017	32.0	7360	0.0061	0.8650	0.9129
0.0017	33.0	7590	0.0062	0.8656	0.9138
0.0017	34.0	7820	0.0064	0.8657	0.9127
0.0016	35.0	8050	0.0065	0.8665	0.9156
0.0016	36.0	8280	0.0067	0.8624	0.9051
0.0015	37.0	8510	0.0065	0.8658	0.9116
0.0016	38.0	8740	0.0061	0.8660	0.9149
0.0015	39.0	8970	0.0063	0.8662	0.9155
0.0015	40.0	9200	0.0062	0.8656	0.9155

Framework versions

• Transformers 4.42.3
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1