segformer-b0-finetuned-segments-pv_v1_3x_normalized_p100_4batch

This model is a fine-tuned version of nvidia/segformer-b0-finetuned-ade-512-512 on the mouadenna/satellite_PV_dataset_train_test_v1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0067
• Mean Iou: 0.8641
• Precision: 0.9173

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: 0.0004
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.001
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Precision
0.0077	0.9993	687	0.0077	0.7897	0.8235
0.0056	2.0	1375	0.0059	0.8193	0.8760
0.0065	2.9993	2062	0.0064	0.8222	0.9068
0.0047	4.0	2750	0.0061	0.8195	0.9299
0.0039	4.9993	3437	0.0055	0.8440	0.9075
0.0044	6.0	4125	0.0063	0.8208	0.8479
0.0034	6.9993	4812	0.0080	0.7750	0.8153
0.0037	8.0	5500	0.0053	0.8475	0.9084
0.004	8.9993	6187	0.0073	0.8013	0.8237
0.003	10.0	6875	0.0056	0.8476	0.8955
0.0038	10.9993	7562	0.0058	0.8273	0.9144
0.0028	12.0	8250	0.0065	0.8143	0.8888
0.0031	12.9993	8937	0.0064	0.8175	0.9188
0.003	14.0	9625	0.0051	0.8491	0.9027
0.0025	14.9993	10312	0.0059	0.8558	0.9085
0.0029	16.0	11000	0.0057	0.8454	0.9029
0.0026	16.9993	11687	0.0057	0.8547	0.9230
0.0024	18.0	12375	0.0059	0.8579	0.9045
0.0025	18.9993	13062	0.0059	0.8645	0.9094
0.0025	20.0	13750	0.0059	0.8498	0.9174
0.0024	20.9993	14437	0.0056	0.8576	0.8970
0.0022	22.0	15125	0.0063	0.8541	0.8952
0.0031	22.9993	15812	0.0054	0.8508	0.9154
0.0021	24.0	16500	0.0057	0.8545	0.9119
0.0022	24.9993	17187	0.0058	0.8474	0.9149
0.0022	26.0	17875	0.0066	0.8325	0.8879
0.0021	26.9993	18562	0.0062	0.8522	0.9156
0.0021	28.0	19250	0.0063	0.8488	0.8932
0.002	28.9993	19937	0.0061	0.8579	0.9200
0.002	30.0	20625	0.0059	0.8624	0.9182
0.0021	30.9993	21312	0.0061	0.8564	0.9013
0.0019	32.0	22000	0.0060	0.8601	0.9091
0.0018	32.9993	22687	0.0059	0.8640	0.9163
0.0017	34.0	23375	0.0062	0.8622	0.9187
0.0017	34.9993	24062	0.0062	0.8634	0.9245
0.0017	36.0	24750	0.0064	0.8655	0.9196
0.0017	36.9993	25437	0.0063	0.8642	0.9197
0.0016	38.0	26125	0.0065	0.8634	0.9166
0.0016	38.9993	26812	0.0067	0.8639	0.9186
0.0016	39.9709	27480	0.0067	0.8641	0.9173

Framework versions

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