segformer-b2-finetuned-segments-pv_v1_normalized_p100_4batch_fp

This model is a fine-tuned version of nvidia/segformer-b2-finetuned-ade-512-512 on the mouadenna/satellite_PV_dataset_train_test_v1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0046
• Mean Iou: 0.8880
• 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: 1e-05
• 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: 0.01
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Precision
0.6668	0.9989	229	0.4009	0.5075	0.5321
0.2583	1.9978	458	0.1436	0.6208	0.6535
0.1355	2.9967	687	0.0809	0.7078	0.7644
0.088	4.0	917	0.0585	0.7472	0.8136
0.0638	4.9989	1146	0.0452	0.7737	0.8353
0.05	5.9978	1375	0.0365	0.7845	0.8394
0.0401	6.9967	1604	0.0344	0.8087	0.8717
0.0332	8.0	1834	0.0277	0.8128	0.8682
0.0286	8.9989	2063	0.0188	0.8210	0.8710
0.0247	9.9978	2292	0.0148	0.8369	0.8881
0.0214	10.9967	2521	0.0133	0.8332	0.8716
0.0189	12.0	2751	0.0156	0.8286	0.8597
0.017	12.9989	2980	0.0139	0.8397	0.8726
0.0151	13.9978	3209	0.0154	0.8544	0.8943
0.0139	14.9967	3438	0.0114	0.8553	0.8897
0.0127	16.0	3668	0.0108	0.8517	0.8799
0.0118	16.9989	3897	0.0075	0.8658	0.9040
0.0108	17.9978	4126	0.0094	0.8700	0.9088
0.0101	18.9967	4355	0.0084	0.8746	0.9151
0.0094	20.0	4585	0.0071	0.8693	0.8973
0.0088	20.9989	4814	0.0071	0.8668	0.8931
0.0082	21.9978	5043	0.0060	0.8786	0.9151
0.008	22.9967	5272	0.0063	0.8776	0.9109
0.0075	24.0	5502	0.0066	0.8776	0.9052
0.0071	24.9989	5731	0.0060	0.8807	0.9115
0.0069	25.9978	5960	0.0062	0.8766	0.9004
0.0065	26.9967	6189	0.0059	0.8754	0.8963
0.0063	28.0	6419	0.0062	0.8825	0.9086
0.006	28.9989	6648	0.0050	0.8839	0.9101
0.0059	29.9978	6877	0.0051	0.8827	0.9069
0.0057	30.9967	7106	0.0056	0.8822	0.9053
0.0055	32.0	7336	0.0047	0.8866	0.9133
0.0055	32.9989	7565	0.0046	0.8876	0.9135
0.0053	33.9978	7794	0.0052	0.8839	0.9053
0.0052	34.9967	8023	0.0048	0.8828	0.9035
0.0051	36.0	8253	0.0046	0.8897	0.9156
0.005	36.9989	8482	0.0045	0.8891	0.9137
0.005	37.9978	8711	0.0047	0.8881	0.9120
0.005	38.9967	8940	0.0047	0.8879	0.9110
0.0049	39.9564	9160	0.0046	0.8880	0.9115

Framework versions

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