segformer-b4-finetuned-UBC

This model is a fine-tuned version of nvidia/segformer-b4-finetuned-ade-512-512 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.9831
• Mean Iou: 0.3535
• Mean Accuracy: 0.5149
• Overall Accuracy: 0.6482
• Accuracy Background: nan
• Accuracy Residential: 0.8665
• Accuracy Commercial: 0.5016
• Accuracy Industrial: 0.5567
• Accuracy Public: 0.4487
• Accuracy Other: 0.2007
• Iou Background: nan
• Iou Residential: 0.7569
• Iou Commercial: 0.3413
• Iou Industrial: 0.2314
• Iou Public: 0.2827
• Iou Other: 0.1552

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: 5
• eval_batch_size: 5
• 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: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Residential	Accuracy Commercial	Accuracy Industrial	Accuracy Public	Accuracy Other	Iou Background	Iou Residential	Iou Commercial	Iou Industrial	Iou Public	Iou Other
0.6613	1.0	112	1.0186	0.2282	0.3923	0.6080	nan	0.7908	0.6591	0.1395	0.3716	0.0005	0.0	0.7023	0.3610	0.0698	0.2355	0.0005
0.6229	2.0	224	0.9143	0.3063	0.4309	0.6547	nan	0.8927	0.4666	0.1872	0.6030	0.0052	nan	0.7356	0.3453	0.0835	0.3617	0.0051
0.474	3.0	336	0.9810	0.3225	0.4892	0.6415	nan	0.8693	0.3502	0.5204	0.6787	0.0273	nan	0.7447	0.2791	0.2138	0.3482	0.0267
0.416	4.0	448	1.0021	0.3657	0.5521	0.6573	nan	0.8784	0.4672	0.7352	0.4826	0.1971	nan	0.7372	0.3384	0.2748	0.3262	0.1518
0.3421	5.0	560	1.0451	0.3722	0.5476	0.6509	nan	0.8677	0.4221	0.6045	0.5012	0.3425	nan	0.7387	0.3117	0.2801	0.3251	0.2053
0.2723	6.0	672	1.2387	0.3482	0.4987	0.6411	nan	0.8439	0.5259	0.4330	0.4398	0.2511	nan	0.7547	0.3376	0.1830	0.2817	0.1840
0.2719	7.0	784	1.1848	0.3489	0.4784	0.6576	nan	0.8764	0.5764	0.3244	0.4185	0.1962	nan	0.7348	0.3760	0.1931	0.2871	0.1537
0.1785	8.0	896	1.2896	0.3686	0.5563	0.6438	nan	0.8694	0.4545	0.7179	0.3925	0.3472	nan	0.7536	0.3200	0.3063	0.2641	0.1988
0.1309	9.0	1008	1.3525	0.3515	0.4856	0.6480	nan	0.8511	0.4882	0.3594	0.5495	0.1796	nan	0.7557	0.3365	0.2199	0.3031	0.1426
0.1106	10.0	1120	1.4989	0.3494	0.5203	0.6379	nan	0.8556	0.3745	0.5852	0.5822	0.2041	nan	0.7496	0.2844	0.2445	0.3171	0.1513
0.1281	11.0	1232	1.5308	0.3652	0.5315	0.6550	nan	0.8469	0.5610	0.5537	0.4390	0.2569	nan	0.7555	0.3650	0.2274	0.3013	0.1770
0.0942	12.0	1344	1.5054	0.3547	0.5125	0.6486	nan	0.8600	0.4925	0.5328	0.4894	0.1876	nan	0.7544	0.3393	0.2390	0.2960	0.1450
0.1254	13.0	1456	1.5499	0.3497	0.5026	0.6400	nan	0.8533	0.5087	0.4680	0.4197	0.2634	nan	0.7487	0.3391	0.2159	0.2656	0.1794
0.1043	14.0	1568	1.5838	0.3543	0.5093	0.6513	nan	0.8712	0.4877	0.4996	0.4779	0.2101	nan	0.7567	0.3410	0.2234	0.2991	0.1514
0.0841	15.0	1680	1.6761	0.3677	0.5503	0.6507	nan	0.8639	0.4700	0.7035	0.4673	0.2470	nan	0.7547	0.3283	0.2825	0.3011	0.1716
0.0765	16.0	1792	1.7000	0.3558	0.5328	0.6526	nan	0.9054	0.3970	0.6789	0.4872	0.1955	nan	0.7523	0.3063	0.2699	0.3055	0.1451
0.0712	17.0	1904	1.8459	0.3587	0.5353	0.6467	nan	0.8638	0.4774	0.6618	0.4534	0.2200	nan	0.7542	0.3283	0.2581	0.2895	0.1635
0.0819	18.0	2016	1.7872	0.3544	0.5048	0.6563	nan	0.8925	0.5231	0.4922	0.4012	0.2150	nan	0.7600	0.3556	0.2259	0.2710	0.1596
0.0622	19.0	2128	1.8449	0.3649	0.5505	0.6451	nan	0.8498	0.4761	0.6958	0.4564	0.2742	nan	0.7556	0.3268	0.2658	0.2930	0.1830
0.0353	20.0	2240	1.8546	0.3511	0.5185	0.6412	nan	0.8546	0.4634	0.5779	0.4835	0.2131	nan	0.7545	0.3182	0.2255	0.2945	0.1626
0.0302	21.0	2352	1.8425	0.3545	0.5222	0.6468	nan	0.8731	0.4495	0.5990	0.4813	0.2082	nan	0.7563	0.3200	0.2444	0.2939	0.1580
0.0523	22.0	2464	1.9309	0.3565	0.5288	0.6450	nan	0.8509	0.4838	0.6172	0.4812	0.2110	nan	0.7561	0.3311	0.2390	0.2947	0.1618
0.0374	23.0	2576	1.9155	0.3529	0.5157	0.6481	nan	0.8688	0.4734	0.5491	0.4763	0.2110	nan	0.7528	0.3317	0.2218	0.2983	0.1597
0.0466	24.0	2688	1.9691	0.3515	0.5178	0.6437	nan	0.8645	0.4620	0.5759	0.4724	0.2140	nan	0.7538	0.3246	0.2296	0.2893	0.1599
0.0243	25.0	2800	1.9683	0.3565	0.5246	0.6461	nan	0.8617	0.4950	0.5975	0.4406	0.2284	nan	0.7571	0.3355	0.2414	0.2802	0.1684
0.0287	26.0	2912	1.9427	0.3552	0.5152	0.6491	nan	0.8651	0.5151	0.5380	0.4333	0.2246	nan	0.7555	0.3468	0.2270	0.2802	0.1668
0.0575	27.0	3024	2.0501	0.3538	0.5194	0.6471	nan	0.8583	0.5049	0.5777	0.4570	0.1992	nan	0.7564	0.3415	0.2293	0.2877	0.1538
0.0523	28.0	3136	1.9851	0.3553	0.5166	0.6489	nan	0.8639	0.5234	0.5550	0.4250	0.2156	nan	0.7565	0.3483	0.2329	0.2759	0.1629
0.0243	29.0	3248	1.9891	0.3533	0.5148	0.6480	nan	0.8676	0.5087	0.5591	0.4325	0.2064	nan	0.7561	0.3427	0.2312	0.2779	0.1584
0.0514	30.0	3360	1.9831	0.3535	0.5149	0.6482	nan	0.8665	0.5016	0.5567	0.4487	0.2007	nan	0.7569	0.3413	0.2314	0.2827	0.1552

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.2