segformer-b5-finetuned-morphpadver1-hgo-coord-v9_mix_resample_40epochs

This model is a fine-tuned version of nvidia/mit-b5 on the NICOPOI-9/morphpad_coord_hgo_512_4class_v2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6422
• Mean Iou: 0.7347
• Mean Accuracy: 0.8436
• Overall Accuracy: 0.8478
• Accuracy 0-0: 0.7840
• Accuracy 0-90: 0.8932
• Accuracy 90-0: 0.8779
• Accuracy 90-90: 0.8194
• Iou 0-0: 0.7200
• Iou 0-90: 0.7429
• Iou 90-0: 0.7413
• Iou 90-90: 0.7348

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: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy 0-0	Accuracy 0-90	Accuracy 90-0	Accuracy 90-90	Iou 0-0	Iou 0-90	Iou 90-0	Iou 90-90
1.3684	1.3638	4000	1.3355	0.1756	0.3310	0.3524	0.1001	0.3549	0.7755	0.0937	0.0866	0.2432	0.2910	0.0815
0.6699	2.7276	8000	1.1446	0.3199	0.4833	0.4927	0.3842	0.5642	0.5994	0.3853	0.2808	0.3636	0.3545	0.2808
0.7075	4.0914	12000	1.0016	0.4314	0.6036	0.6040	0.6056	0.5476	0.6773	0.5840	0.4196	0.4287	0.4540	0.4233
0.4978	5.4552	16000	0.8600	0.5036	0.6641	0.6709	0.5657	0.7738	0.6865	0.6305	0.4895	0.5093	0.5085	0.5071
0.6194	6.8190	20000	0.8307	0.5419	0.7019	0.7030	0.6992	0.7035	0.7266	0.6782	0.5500	0.5426	0.5441	0.5311
0.2622	8.1827	24000	0.7177	0.5987	0.7469	0.7491	0.7321	0.7660	0.7749	0.7148	0.5962	0.5985	0.6019	0.5984
0.9683	9.5465	28000	0.7541	0.5951	0.7409	0.7474	0.6970	0.7901	0.8298	0.6467	0.5996	0.6063	0.6010	0.5736
0.2542	10.9103	32000	0.7039	0.6381	0.7768	0.7791	0.7554	0.7818	0.8233	0.7465	0.6333	0.6498	0.6295	0.6400
0.1691	12.2741	36000	0.6232	0.6632	0.7959	0.7978	0.7673	0.8280	0.8005	0.7877	0.6636	0.6778	0.6536	0.6580
0.1883	13.6379	40000	0.6711	0.6649	0.7948	0.7996	0.7315	0.8565	0.8291	0.7622	0.6514	0.6774	0.6675	0.6632
0.164	15.0017	44000	0.6627	0.6688	0.7980	0.8022	0.7670	0.8227	0.8637	0.7386	0.6740	0.6846	0.6687	0.6479
0.2406	16.3655	48000	0.6364	0.6930	0.8159	0.8194	0.7843	0.8565	0.8466	0.7763	0.6894	0.7005	0.7017	0.6805
0.109	17.7293	52000	0.6087	0.6872	0.8119	0.8153	0.7622	0.8473	0.8443	0.7940	0.6733	0.7055	0.6834	0.6868
0.1262	19.0931	56000	0.6101	0.6999	0.8202	0.8240	0.7795	0.8572	0.8619	0.7823	0.6912	0.7071	0.7041	0.6972
0.1633	20.4569	60000	0.6434	0.7056	0.8239	0.8280	0.7832	0.8548	0.8795	0.7781	0.7006	0.7194	0.7056	0.6969
8.0069	21.8207	64000	0.5640	0.7111	0.8286	0.8319	0.8192	0.8718	0.8567	0.7666	0.7149	0.7267	0.7119	0.6909
0.0335	23.1845	68000	0.5820	0.7215	0.8348	0.8388	0.7894	0.8701	0.8828	0.7967	0.7085	0.7253	0.7293	0.7230
0.3274	24.5482	72000	0.6041	0.7210	0.8346	0.8386	0.7929	0.8767	0.8754	0.7933	0.7157	0.7344	0.7223	0.7117
0.137	25.9120	76000	0.6174	0.7000	0.8197	0.8246	0.7584	0.8768	0.8609	0.7829	0.6897	0.7138	0.7079	0.6886
0.0973	27.2758	80000	0.6329	0.7039	0.8208	0.8276	0.7483	0.9023	0.8783	0.7542	0.6896	0.7184	0.7139	0.6936
0.0938	28.6396	84000	0.5952	0.7273	0.8412	0.8421	0.8351	0.8486	0.8535	0.8276	0.7316	0.7346	0.7211	0.7218
0.0558	30.0034	88000	0.6204	0.7017	0.8193	0.8260	0.7578	0.8958	0.8822	0.7412	0.6987	0.7170	0.7090	0.6822
0.0048	31.3672	92000	0.6403	0.7057	0.8219	0.8289	0.7465	0.8980	0.8915	0.7514	0.6933	0.7244	0.7130	0.6920
0.0134	32.7310	96000	0.6758	0.7192	0.8333	0.8375	0.8073	0.8800	0.8754	0.7704	0.7150	0.7309	0.7256	0.7052
0.0326	34.0948	100000	0.6023	0.7256	0.8362	0.8419	0.7617	0.9030	0.8856	0.7944	0.7094	0.7337	0.7341	0.7254
0.0379	35.4586	104000	0.6208	0.7342	0.8436	0.8472	0.7932	0.8949	0.8641	0.8220	0.7245	0.7368	0.7419	0.7334
0.0676	36.8224	108000	0.6448	0.7298	0.8401	0.8446	0.7877	0.8896	0.8831	0.8001	0.7180	0.7440	0.7336	0.7236
0.0604	38.1862	112000	0.6608	0.7334	0.8426	0.8468	0.7867	0.8845	0.8869	0.8124	0.7214	0.7441	0.7364	0.7315
0.018	39.5499	116000	0.6422	0.7347	0.8436	0.8478	0.7840	0.8932	0.8779	0.8194	0.7200	0.7429	0.7413	0.7348

Framework versions

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