segformer-b1-finetuned-segments-pv_v1_3x_normalized_p100_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: 0.0087
• Mean Iou: 0.8602
• Precision: 0.9152

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
0.0073	0.9993	687	0.0072	0.7997	0.8395
0.0052	2.0	1375	0.0069	0.8039	0.8906
0.0051	2.9993	2062	0.0060	0.8301	0.8951
0.0048	4.0	2750	0.0057	0.8223	0.9070
0.0039	4.9993	3437	0.0054	0.8433	0.9104
0.0042	6.0	4125	0.0054	0.8414	0.8779
0.0031	6.9993	4812	0.0052	0.8453	0.8852
0.0034	8.0	5500	0.0051	0.8526	0.9146
0.0036	8.9993	6187	0.0059	0.8319	0.8884
0.0027	10.0	6875	0.0058	0.8453	0.8990
0.0028	10.9993	7562	0.0052	0.8552	0.9152
0.0027	12.0	8250	0.0062	0.8459	0.9038
0.0032	12.9993	8937	0.0056	0.8506	0.9163
0.0024	14.0	9625	0.0062	0.8529	0.9189
0.0035	14.9993	10312	0.0058	0.8464	0.9102
0.0024	16.0	11000	0.0059	0.8575	0.9126
0.0023	16.9993	11687	0.0057	0.8527	0.9201
0.0024	18.0	12375	0.0060	0.8573	0.9177
0.0028	18.9993	13062	0.0063	0.8601	0.9064
0.0023	20.0	13750	0.0061	0.8589	0.9164
0.002	20.9993	14437	0.0061	0.8611	0.9046
0.002	22.0	15125	0.0057	0.8633	0.9143
0.002	22.9993	15812	0.0067	0.8552	0.9133
0.0018	24.0	16500	0.0068	0.8594	0.9174
0.0021	24.9993	17187	0.0063	0.8545	0.9111
0.0023	26.0	17875	0.0055	0.8642	0.9149
0.0019	26.9993	18562	0.0060	0.8627	0.9152
0.0017	28.0	19250	0.0063	0.8658	0.9148
0.0017	28.9993	19937	0.0067	0.8644	0.9085
0.0017	30.0	20625	0.0068	0.8578	0.9110
0.0017	30.9993	21312	0.0067	0.8585	0.9130
0.0015	32.0	22000	0.0069	0.8613	0.9103
0.0015	32.9993	22687	0.0073	0.8599	0.9200
0.0014	34.0	23375	0.0074	0.8605	0.9181
0.0014	34.9993	24062	0.0079	0.8581	0.9174
0.0013	36.0	24750	0.0081	0.8582	0.9123
0.0013	36.9993	25437	0.0084	0.8599	0.9166
0.0012	38.0	26125	0.0084	0.8603	0.9139
0.0013	38.9993	26812	0.0092	0.8599	0.9193
0.0012	39.9709	27480	0.0087	0.8602	0.9152

Framework versions

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