segformer-b1-finetuned-segments-pv_v1_normalized_t4_4batch

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: 0.0107
• Mean Iou: 0.8767
• Precision: 0.9236

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Precision
0.0061	0.9935	114	0.0077	0.7881	0.8686
0.0043	1.9956	229	0.0063	0.8084	0.8541
0.0051	2.9978	344	0.0052	0.8338	0.9445
0.0141	4.0	459	0.0054	0.8421	0.8811
0.0043	4.9935	573	0.0052	0.8414	0.9099
0.0027	5.9956	688	0.0057	0.8450	0.8967
0.0054	6.9978	803	0.0052	0.8505	0.9364
0.0024	8.0	918	0.0064	0.8408	0.9001
0.0024	8.9935	1032	0.0068	0.8378	0.9158
0.0033	9.9956	1147	0.0055	0.8643	0.9156
0.0014	10.9978	1262	0.0055	0.8611	0.9048
0.0019	12.0	1377	0.0070	0.8410	0.8900
0.0025	12.9935	1491	0.0060	0.8629	0.9112
0.0018	13.9956	1606	0.0063	0.8577	0.9294
0.002	14.9978	1721	0.0063	0.8539	0.8888
0.002	16.0	1836	0.0072	0.8598	0.9172
0.0021	16.9935	1950	0.0062	0.8555	0.9074
0.0018	17.9956	2065	0.0069	0.8598	0.9167
0.0018	18.9978	2180	0.0074	0.8556	0.9160
0.002	20.0	2295	0.0067	0.8662	0.9117
0.003	20.9935	2409	0.0062	0.8724	0.9245
0.0027	21.9956	2524	0.0067	0.8727	0.9124
0.0013	22.9978	2639	0.0068	0.8684	0.9147
0.0011	24.0	2754	0.0070	0.8723	0.9165
0.0014	24.9935	2868	0.0074	0.8709	0.9257
0.0011	25.9956	2983	0.0075	0.8697	0.9139
0.001	26.9978	3098	0.0071	0.8780	0.9273
0.0011	28.0	3213	0.0075	0.8743	0.9182
0.0008	28.9935	3327	0.0080	0.8744	0.9234
0.0007	29.9956	3442	0.0086	0.8692	0.9205
0.001	30.9978	3557	0.0083	0.8720	0.9145
0.0009	32.0	3672	0.0084	0.8745	0.9167
0.0009	32.9935	3786	0.0084	0.8717	0.9155
0.0011	33.9956	3901	0.0084	0.8756	0.9279
0.0007	34.9978	4016	0.0090	0.8777	0.9233
0.0008	36.0	4131	0.0090	0.8744	0.9173
0.0011	36.9935	4245	0.0097	0.8753	0.9192
0.0008	37.9956	4360	0.0091	0.8757	0.9260
0.0009	38.9978	4475	0.0091	0.8739	0.9173
0.0008	40.0	4590	0.0103	0.8760	0.9274
0.0008	40.9935	4704	0.0106	0.8749	0.9263
0.0008	41.9956	4819	0.0097	0.8753	0.9238
0.0009	42.9978	4934	0.0099	0.8730	0.9159
0.0006	44.0	5049	0.0101	0.8757	0.9247
0.0006	44.9935	5163	0.0104	0.8756	0.9217
0.0007	45.9956	5278	0.0106	0.8720	0.9175
0.0006	46.9978	5393	0.0107	0.8753	0.9202
0.0005	48.0	5508	0.0107	0.8757	0.9224
0.0007	48.9935	5622	0.0107	0.8764	0.9227
0.0008	49.6732	5700	0.0107	0.8767	0.9236

Framework versions

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