segformer-b5-finetuned-ade20k-hgo-coord_40epochs_distortion_ver2_global_norm_with_void_3

This model is a fine-tuned version of NICOPOI-9/segformer-b5-finetuned-ade20k-hgo-coord_40epochs_distortion_ver2_global_norm_with_void on the NICOPOI-9/Modphad_Perlin_two_void_coord_global_norm dataset. It achieves the following results on the evaluation set:

• Loss: 0.5837
• Mean Iou: 0.7559
• Mean Accuracy: 0.8600
• Overall Accuracy: 0.8723
• Accuracy [0,0]: 0.8245
• Accuracy [0,1]: 0.8968
• Accuracy [1,0]: 0.9052
• Accuracy [1,1]: 0.8873
• Accuracy [0,2]: 0.8720
• Accuracy [0,3]: 0.8555
• Accuracy [1,2]: 0.8630
• Accuracy [1,3]: 0.9183
• Accuracy [2,0]: 0.8100
• Accuracy [2,1]: 0.8301
• Accuracy [2,2]: 0.8209
• Accuracy [2,3]: 0.8314
• Accuracy [3,0]: 0.8513
• Accuracy [3,1]: 0.7988
• Accuracy [3,2]: 0.8444
• Accuracy [3,3]: 0.8538
• Accuracy Void: 0.9565
• Iou [0,0]: 0.7600
• Iou [0,1]: 0.7777
• Iou [1,0]: 0.7713
• Iou [1,1]: 0.7672
• Iou [0,2]: 0.7510
• Iou [0,3]: 0.7562
• Iou [1,2]: 0.7433
• Iou [1,3]: 0.8024
• Iou [2,0]: 0.7175
• Iou [2,1]: 0.7317
• Iou [2,2]: 0.6962
• Iou [2,3]: 0.7339
• Iou [3,0]: 0.7656
• Iou [3,1]: 0.7285
• Iou [3,2]: 0.7081
• Iou [3,3]: 0.7214
• Iou Void: 0.9178

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: 6e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 80

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy [0,0]	Accuracy [0,1]	Accuracy [1,0]	Accuracy [1,1]	Accuracy [0,2]	Accuracy [0,3]	Accuracy [1,2]	Accuracy [1,3]	Accuracy [2,0]	Accuracy [2,1]	Accuracy [2,2]	Accuracy [2,3]	Accuracy [3,0]	Accuracy [3,1]	Accuracy [3,2]	Accuracy [3,3]	Accuracy Void	Iou [0,0]	Iou [0,1]	Iou [1,0]	Iou [1,1]	Iou [0,2]	Iou [0,3]	Iou [1,2]	Iou [1,3]	Iou [2,0]	Iou [2,1]	Iou [2,2]	Iou [2,3]	Iou [3,0]	Iou [3,1]	Iou [3,2]	Iou [3,3]	Iou Void
0.152	7.3260	4000	0.7633	0.6318	0.7756	0.7910	0.7271	0.7253	0.8081	0.8546	0.8294	0.7726	0.7022	0.8399	0.6781	0.7753	0.7977	0.7897	0.8184	0.5306	0.7908	0.8273	0.9181	0.6549	0.6477	0.6603	0.6563	0.6384	0.6154	0.6197	0.6871	0.5669	0.6722	0.5917	0.6063	0.6417	0.4889	0.5758	0.5383	0.8792
0.4745	14.6520	8000	0.6847	0.6572	0.7918	0.8081	0.7585	0.7823	0.8711	0.8361	0.8139	0.7747	0.6589	0.8644	0.7717	0.8130	0.7244	0.7572	0.8354	0.6652	0.7897	0.8170	0.9266	0.6768	0.6619	0.6854	0.6303	0.6853	0.6383	0.6093	0.6536	0.6602	0.5989	0.6355	0.6343	0.6964	0.6043	0.6389	0.5706	0.8930
0.2779	21.9780	12000	0.6080	0.6946	0.8203	0.8326	0.7822	0.8660	0.8674	0.8758	0.8260	0.8204	0.8000	0.8680	0.7232	0.7985	0.8629	0.7841	0.8052	0.7116	0.8226	0.8120	0.9190	0.7199	0.7346	0.7254	0.6937	0.6854	0.6604	0.6958	0.7238	0.6568	0.6896	0.6128	0.6912	0.7207	0.6371	0.6627	0.6076	0.8904
0.0837	29.3040	16000	0.5950	0.7067	0.8279	0.8404	0.7697	0.8721	0.8696	0.8884	0.8008	0.8170	0.7844	0.8911	0.8288	0.7954	0.7810	0.7962	0.8401	0.7972	0.7915	0.8268	0.9235	0.6886	0.7537	0.6917	0.7075	0.7107	0.6793	0.6946	0.7357	0.7132	0.6818	0.6766	0.7030	0.6825	0.6735	0.6497	0.6747	0.8970
0.1192	36.6300	20000	0.6074	0.7169	0.8348	0.8475	0.8083	0.8478	0.8681	0.8823	0.8403	0.8282	0.8310	0.9117	0.7915	0.8006	0.7987	0.7866	0.8475	0.7392	0.8417	0.8336	0.9345	0.7405	0.7410	0.7164	0.7271	0.7169	0.6911	0.7208	0.7466	0.6942	0.7163	0.6798	0.6821	0.7093	0.6520	0.6803	0.6701	0.9026
0.0538	43.9560	24000	0.6011	0.7190	0.8335	0.8496	0.8453	0.8891	0.8780	0.8929	0.8250	0.7916	0.7874	0.8868	0.8056	0.8231	0.7849	0.7742	0.8565	0.7412	0.7993	0.8380	0.9511	0.7570	0.6969	0.7359	0.7488	0.7296	0.6807	0.6977	0.7651	0.7055	0.6898	0.6898	0.6849	0.7412	0.6480	0.6828	0.6554	0.9143
0.0616	51.2821	28000	0.5992	0.7349	0.8451	0.8586	0.8345	0.9088	0.8899	0.8821	0.8455	0.8010	0.8537	0.8853	0.7653	0.8475	0.8179	0.8252	0.8186	0.7725	0.8359	0.8324	0.9512	0.7540	0.7328	0.7120	0.7596	0.7493	0.7131	0.7134	0.7649	0.6818	0.7506	0.7135	0.7105	0.7632	0.7053	0.6837	0.6767	0.9098
0.0471	58.6081	32000	0.5636	0.7472	0.8530	0.8671	0.8347	0.8966	0.9209	0.8917	0.8560	0.8639	0.8388	0.9071	0.7870	0.8462	0.8135	0.8072	0.8454	0.7765	0.8370	0.8212	0.9580	0.7641	0.7822	0.7606	0.7573	0.7466	0.7331	0.7421	0.7789	0.6958	0.7347	0.6971	0.7076	0.7581	0.7122	0.7289	0.6887	0.9149
0.101	65.9341	36000	0.5694	0.7514	0.8573	0.8696	0.8289	0.8963	0.8971	0.8826	0.8476	0.8269	0.8581	0.9075	0.8128	0.8565	0.8462	0.8263	0.8474	0.8111	0.8374	0.8324	0.9600	0.7687	0.7604	0.7533	0.7885	0.7427	0.7313	0.7500	0.7977	0.7090	0.7323	0.6802	0.7167	0.7713	0.7278	0.7082	0.7185	0.9169
0.0344	73.2601	40000	0.5837	0.7559	0.8600	0.8723	0.8245	0.8968	0.9052	0.8873	0.8720	0.8555	0.8630	0.9183	0.8100	0.8301	0.8209	0.8314	0.8513	0.7988	0.8444	0.8538	0.9565	0.7600	0.7777	0.7713	0.7672	0.7510	0.7562	0.7433	0.8024	0.7175	0.7317	0.6962	0.7339	0.7656	0.7285	0.7081	0.7214	0.9178

Framework versions

• Transformers 4.48.3
• Pytorch 2.1.0
• Datasets 3.2.0
• Tokenizers 0.21.0