segformer_b2_finetuned_segment_pv_p100_4batch

This model is a fine-tuned version of nvidia/segformer-b2-finetuned-ade-512-512 on the mouadenna/satellite_PV_dataset_train_test_v1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0090
• Mean Iou: 0.8765
• Precision: 0.9192

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: 4e-05
• train_batch_size: 4
• eval_batch_size: 4
• 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
0.5015	1.0	917	0.1494	0.5660	0.6026
0.0714	2.0	1834	0.0237	0.7528	0.7988
0.0171	3.0	2751	0.0101	0.7978	0.8930
0.0087	4.0	3668	0.0072	0.8260	0.8534
0.0058	5.0	4585	0.0067	0.8418	0.8981
0.0046	6.0	5502	0.0056	0.8457	0.8971
0.0038	7.0	6419	0.0056	0.8530	0.8770
0.0034	8.0	7336	0.0056	0.8525	0.8978
0.003	9.0	8253	0.0052	0.8643	0.9063
0.0028	10.0	9170	0.0054	0.8641	0.9010
0.0027	11.0	10087	0.0065	0.8489	0.9236
0.0025	12.0	11004	0.0066	0.8432	0.9006
0.0024	13.0	11921	0.0055	0.8637	0.9242
0.0022	14.0	12838	0.0054	0.8679	0.9104
0.0024	15.0	13755	0.0055	0.8719	0.9171
0.0019	16.0	14672	0.0055	0.8746	0.9219
0.0019	17.0	15589	0.0056	0.8668	0.9062
0.0018	18.0	16506	0.0063	0.8703	0.9121
0.0017	19.0	17423	0.0062	0.8694	0.9084
0.0016	20.0	18340	0.0063	0.8719	0.9133
0.0015	21.0	19257	0.0065	0.8734	0.9159
0.0014	22.0	20174	0.0068	0.8730	0.9155
0.0015	23.0	21091	0.0069	0.8719	0.9228
0.0013	24.0	22008	0.0069	0.8745	0.9162
0.0013	25.0	22925	0.0069	0.8757	0.9196
0.0012	26.0	23842	0.0075	0.8747	0.9138
0.0012	27.0	24759	0.0074	0.8750	0.9159
0.0012	28.0	25676	0.0074	0.8755	0.9213
0.0011	29.0	26593	0.0081	0.8762	0.9154
0.0011	30.0	27510	0.0083	0.8754	0.9162
0.0011	31.0	28427	0.0084	0.8753	0.9168
0.001	32.0	29344	0.0083	0.8754	0.9202
0.001	33.0	30261	0.0085	0.8758	0.9174
0.001	34.0	31178	0.0085	0.8758	0.9208
0.0009	35.0	32095	0.0088	0.8763	0.9191
0.0009	36.0	33012	0.0090	0.8756	0.9172
0.0009	37.0	33929	0.0090	0.8760	0.9181
0.0009	38.0	34846	0.0087	0.8764	0.9195
0.0009	39.0	35763	0.0090	0.8763	0.9184
0.0009	40.0	36680	0.0090	0.8765	0.9192

Framework versions

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