segformer-b0-finetuned-segments-pv_v1_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: 0.0074
• Mean Iou: 0.8483
• Precision: 0.9169

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
0.0127	0.9989	229	0.0092	0.7982	0.8641
0.0077	1.9978	458	0.0094	0.7871	0.8456
0.006	2.9967	687	0.0067	0.8140	0.9089
0.0051	4.0	917	0.0058	0.8358	0.8713
0.0045	4.9989	1146	0.0059	0.8258	0.8761
0.0042	5.9978	1375	0.0058	0.8415	0.9018
0.0036	6.9967	1604	0.0051	0.8513	0.9049
0.0038	8.0	1834	0.0062	0.8226	0.9256
0.004	8.9989	2063	0.0057	0.8358	0.8913
0.0035	9.9978	2292	0.0053	0.8485	0.9079
0.0037	10.9967	2521	0.0059	0.8192	0.9056
0.0038	12.0	2751	0.0054	0.8487	0.8921
0.0033	12.9989	2980	0.0053	0.8541	0.9086
0.0028	13.9978	3209	0.0055	0.8551	0.8985
0.0026	14.9967	3438	0.0060	0.8483	0.9085
0.0026	16.0	3668	0.0057	0.8495	0.9076
0.0024	16.9989	3897	0.0058	0.8442	0.9083
0.0038	17.9978	4126	0.0066	0.8113	0.8910
0.0031	18.9967	4355	0.0062	0.8488	0.9108
0.0026	20.0	4585	0.0058	0.8575	0.9126
0.0024	20.9989	4814	0.0057	0.8580	0.9119
0.0025	21.9978	5043	0.0059	0.8505	0.8957
0.0031	22.9967	5272	0.0062	0.8472	0.9135
0.0022	24.0	5502	0.0055	0.8598	0.9147
0.0023	24.9989	5731	0.0058	0.8621	0.9090
0.0023	25.9978	5960	0.0064	0.8498	0.9094
0.0023	26.9967	6189	0.0067	0.8428	0.9137
0.0021	28.0	6419	0.0063	0.8527	0.9076
0.002	28.9989	6648	0.0065	0.8509	0.9187
0.002	29.9978	6877	0.0074	0.8424	0.9179
0.002	30.9967	7106	0.0065	0.8577	0.9116
0.0019	32.0	7336	0.0067	0.8547	0.9141
0.0019	32.9989	7565	0.0072	0.8519	0.9168
0.0019	33.9978	7794	0.0067	0.8569	0.9148
0.0019	34.9967	8023	0.0070	0.8544	0.9139
0.0017	36.0	8253	0.0072	0.8510	0.9124
0.0018	36.9989	8482	0.0081	0.8425	0.9164
0.0017	37.9978	8711	0.0073	0.8512	0.9155
0.0018	38.9967	8940	0.0073	0.8495	0.9164
0.0018	39.9564	9160	0.0074	0.8483	0.9169

Framework versions

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