segformer-b1-finetuned-segments-pv_v1_x3_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: 0.0064
• Mean Iou: 0.8466
• Precision: 0.9220

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
0.0084	0.9993	687	0.0063	0.8160	0.8736
0.007	2.0	1375	0.0060	0.8262	0.9006
0.006	2.9993	2062	0.0066	0.8072	0.9214
0.0049	4.0	2750	0.0054	0.8283	0.9287
0.004	4.9993	3437	0.0070	0.8326	0.9068
0.0042	6.0	4125	0.0053	0.8318	0.8834
0.004	6.9993	4812	0.0053	0.8370	0.8893
0.0037	8.0	5500	0.0075	0.8049	0.9404
0.0036	8.9993	6187	0.0074	0.8222	0.9106
0.0033	10.0	6875	0.0061	0.8297	0.9161
0.0031	10.9993	7562	0.0055	0.8427	0.9086
0.0033	12.0	8250	0.0052	0.8437	0.9152
0.0037	12.9993	8937	0.0055	0.8387	0.9186
0.0028	14.0	9625	0.0060	0.8416	0.9137
0.0027	14.9993	10312	0.0052	0.8489	0.9212
0.003	16.0	11000	0.0065	0.8393	0.9158
0.0025	16.9993	11687	0.0063	0.8347	0.9245
0.0027	18.0	12375	0.0065	0.8439	0.9093
0.0032	18.9993	13062	0.0056	0.8495	0.9186
0.0024	20.0	13750	0.0064	0.8466	0.9220

Framework versions

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