segformer_b0_finetuned_segment_pv_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: 0.0065
• Mean Iou: 0.8630
• Precision: 0.9115

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: 4e-05
• train_batch_size: 4
• eval_batch_size: 4
• 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
0.5732	1.0	917	0.2901	0.4717	0.5866
0.1327	2.0	1834	0.0327	0.6919	0.7689
0.0272	3.0	2751	0.0138	0.7618	0.8768
0.0128	4.0	3668	0.0098	0.7875	0.8206
0.0081	5.0	4585	0.0077	0.8165	0.8512
0.0061	6.0	5502	0.0071	0.8177	0.8795
0.005	7.0	6419	0.0060	0.8303	0.8747
0.0045	8.0	7336	0.0056	0.8459	0.8897
0.004	9.0	8253	0.0057	0.8470	0.8851
0.0038	10.0	9170	0.0058	0.8384	0.8761
0.0034	11.0	10087	0.0056	0.8495	0.8966
0.0033	12.0	11004	0.0053	0.8464	0.8956
0.0031	13.0	11921	0.0060	0.8354	0.8843
0.003	14.0	12838	0.0063	0.8414	0.8897
0.0028	15.0	13755	0.0062	0.8466	0.9129
0.0029	16.0	14672	0.0060	0.8480	0.9057
0.0026	17.0	15589	0.0056	0.8559	0.9005
0.0027	18.0	16506	0.0055	0.8571	0.9042
0.0025	19.0	17423	0.0056	0.8571	0.9096
0.0025	20.0	18340	0.0080	0.8329	0.9194
0.0025	21.0	19257	0.0058	0.8567	0.8981
0.0023	22.0	20174	0.0058	0.8624	0.9061
0.0023	23.0	21091	0.0059	0.8599	0.9055
0.0022	24.0	22008	0.0061	0.8601	0.9132
0.0023	25.0	22925	0.0059	0.8603	0.9007
0.0021	26.0	23842	0.0065	0.8594	0.9160
0.0021	27.0	24759	0.0059	0.8636	0.9071
0.0021	28.0	25676	0.0060	0.8650	0.9093
0.002	29.0	26593	0.0061	0.8639	0.9158
0.002	30.0	27510	0.0063	0.8621	0.9074
0.002	31.0	28427	0.0064	0.8598	0.9081
0.0021	32.0	29344	0.0064	0.8570	0.9129
0.0019	33.0	30261	0.0064	0.8601	0.9086
0.0019	34.0	31178	0.0062	0.8626	0.9146
0.0019	35.0	32095	0.0066	0.8607	0.9060
0.0018	36.0	33012	0.0064	0.8610	0.9056
0.0018	37.0	33929	0.0065	0.8618	0.9072
0.0018	38.0	34846	0.0063	0.8631	0.9094
0.0018	39.0	35763	0.0064	0.8628	0.9126
0.0018	40.0	36680	0.0065	0.8630	0.9115

Framework versions

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