segformer_b1_finetuned_segment_pv_p100_32batch

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: 0.0059
• Mean Iou: 0.8651
• Precision: 0.9226

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: 0.00032
• train_batch_size: 32
• eval_batch_size: 32
• 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
0.4433	1.0	115	0.1593	0.5991	0.6512
0.0857	2.0	230	0.0270	0.7731	0.8493
0.0201	3.0	345	0.0116	0.8058	0.8974
0.0108	4.0	460	0.0084	0.8146	0.8758
0.0077	5.0	575	0.0070	0.8267	0.8871
0.0059	6.0	690	0.0072	0.8241	0.9128
0.0051	7.0	805	0.0058	0.8433	0.9197
0.0044	8.0	920	0.0059	0.8466	0.8994
0.0042	9.0	1035	0.0055	0.8474	0.9075
0.0037	10.0	1150	0.0054	0.8576	0.9100
0.0033	11.0	1265	0.0056	0.8555	0.9254
0.0032	12.0	1380	0.0059	0.8455	0.8795
0.0032	13.0	1495	0.0055	0.8600	0.9226
0.0033	14.0	1610	0.0057	0.8558	0.9234
0.0029	15.0	1725	0.0063	0.8533	0.9211
0.003	16.0	1840	0.0072	0.8498	0.9261
0.0035	17.0	1955	0.0102	0.7815	0.9482
0.0033	18.0	2070	0.0244	0.5662	0.9688
0.0028	19.0	2185	0.0256	0.5643	0.9675
0.0027	20.0	2300	0.0078	0.8405	0.9370
0.0026	21.0	2415	0.0241	0.6404	0.9706
0.0024	22.0	2530	0.0492	0.3084	0.9702
0.0025	23.0	2645	0.1065	0.0107	0.9109
0.0024	24.0	2760	0.0958	0.0374	0.8273
0.003	25.0	2875	0.0571	0.1779	0.9912
0.0026	26.0	2990	0.0968	0.0140	0.9839
0.0023	27.0	3105	0.0454	0.2833	0.9702
0.0022	28.0	3220	0.0519	0.2828	0.9658
0.0021	29.0	3335	0.0446	0.3157	0.9698
0.002	30.0	3450	0.0415	0.3630	0.9702
0.002	31.0	3565	0.0308	0.4995	0.9737
0.002	32.0	3680	0.0227	0.6260	0.9700
0.0019	33.0	3795	0.0131	0.7631	0.9631
0.0019	34.0	3910	0.0102	0.8131	0.9541
0.0021	35.0	4025	0.0058	0.8450	0.9449
0.0018	36.0	4140	0.0073	0.8556	0.9326
0.0019	37.0	4255	0.0060	0.8601	0.9339
0.0018	38.0	4370	0.0060	0.8654	0.9244
0.0019	39.0	4485	0.0061	0.8636	0.9248
0.0017	40.0	4600	0.0059	0.8651	0.9226

Framework versions

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