upernet-convnext-tiny-segments-GFB

This model is a fine-tuned version of openmmlab/upernet-convnext-tiny on the segments/GFB dataset. It achieves the following results on the evaluation set:

• Loss: 0.7882
• Mean Iou: 0.7125
• Mean Accuracy: 0.8017
• Overall Accuracy: 0.9250
• Accuracy Unlabeled: 0.9689
• Accuracy Gbm: 0.8241
• Accuracy Podo: 0.7558
• Accuracy Endo: 0.6579
• Iou Unlabeled: 0.9223
• Iou Gbm: 0.7200
• Iou Podo: 0.6457
• Iou Endo: 0.5619

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Gbm	Accuracy Podo	Accuracy Endo	Iou Unlabeled	Iou Gbm	Iou Podo	Iou Endo
0.3378	1.0989	100	0.3889	0.6226	0.7114	0.9018	0.9706	0.6929	0.7052	0.4767	0.9011	0.6333	0.5622	0.3936
0.1929	2.1978	200	0.3725	0.6386	0.7142	0.9080	0.9753	0.7590	0.6493	0.4734	0.9058	0.6620	0.5665	0.4202
0.211	3.2967	300	0.3448	0.6794	0.7742	0.9157	0.9732	0.7392	0.7140	0.6704	0.9153	0.6771	0.5943	0.5310
0.1732	4.3956	400	0.3507	0.6862	0.8303	0.9114	0.9322	0.9151	0.8061	0.6677	0.9085	0.6941	0.6330	0.5094
0.1567	5.4945	500	0.3476	0.7074	0.8276	0.9220	0.9573	0.8203	0.8055	0.7274	0.9211	0.7174	0.6420	0.5491
0.148	6.5934	600	0.3502	0.7144	0.8109	0.9260	0.9657	0.8467	0.7628	0.6683	0.9245	0.7288	0.6449	0.5595
0.1983	7.6923	700	0.3575	0.7204	0.8191	0.9270	0.9651	0.8479	0.7686	0.6948	0.9248	0.7348	0.6498	0.5723
0.1115	8.7912	800	0.4055	0.7153	0.8242	0.9248	0.9591	0.8677	0.7683	0.7017	0.9233	0.7252	0.6469	0.5656
0.0935	9.8901	900	0.3961	0.7268	0.8366	0.9275	0.9613	0.8498	0.7884	0.7471	0.9255	0.7353	0.6576	0.5887
0.1255	10.9890	1000	0.4276	0.7251	0.8230	0.9280	0.9660	0.8403	0.7779	0.7078	0.9256	0.7352	0.6567	0.5830
0.1118	12.0879	1100	0.4431	0.7234	0.8176	0.9282	0.9653	0.8538	0.7817	0.6697	0.9259	0.7353	0.6600	0.5722
0.0811	13.1868	1200	0.4577	0.7245	0.8213	0.9281	0.9647	0.8568	0.7774	0.6863	0.9258	0.7373	0.6566	0.5783
0.0783	14.2857	1300	0.5041	0.7215	0.8105	0.9278	0.9688	0.8259	0.7848	0.6624	0.9250	0.7308	0.6591	0.5713
0.0984	15.3846	1400	0.5062	0.7261	0.8201	0.9288	0.9668	0.8508	0.7722	0.6909	0.9266	0.7359	0.6590	0.5828
0.0755	16.4835	1500	0.5262	0.7257	0.8279	0.9276	0.9628	0.8513	0.7876	0.7098	0.9253	0.7320	0.6609	0.5848
0.0824	17.5824	1600	0.5320	0.7196	0.8075	0.9276	0.9680	0.8450	0.7707	0.6461	0.9250	0.7322	0.6556	0.5659
0.0657	18.6813	1700	0.5438	0.7216	0.8137	0.9277	0.9682	0.8346	0.7714	0.6806	0.9253	0.7305	0.6563	0.5744
0.0808	19.7802	1800	0.5700	0.7205	0.8109	0.9273	0.9688	0.8262	0.7749	0.6736	0.9245	0.7297	0.6554	0.5725
0.0602	20.8791	1900	0.5682	0.7237	0.8168	0.9278	0.9673	0.8404	0.7699	0.6894	0.9251	0.7325	0.6563	0.5809
0.0621	21.9780	2000	0.5880	0.7230	0.8190	0.9273	0.9668	0.8393	0.7647	0.7053	0.9248	0.7297	0.6538	0.5838
0.0691	23.0769	2100	0.6024	0.7189	0.8108	0.9267	0.9690	0.8110	0.7842	0.6790	0.9243	0.7248	0.6538	0.5725
0.0628	24.1758	2200	0.6413	0.7197	0.8077	0.9272	0.9688	0.8354	0.7672	0.6595	0.9243	0.7294	0.6547	0.5705
0.072	25.2747	2300	0.6427	0.7162	0.8039	0.9264	0.9694	0.8201	0.7741	0.6521	0.9235	0.7255	0.6531	0.5626
0.0699	26.3736	2400	0.6491	0.7185	0.8066	0.9270	0.9691	0.8330	0.7650	0.6595	0.9243	0.7280	0.6527	0.5691
0.0666	27.4725	2500	0.6501	0.7213	0.8158	0.9268	0.9669	0.8289	0.7778	0.6897	0.9241	0.7276	0.6548	0.5788
0.0512	28.5714	2600	0.6661	0.7182	0.8072	0.9269	0.9685	0.8448	0.7528	0.6628	0.9244	0.7297	0.6496	0.5693
0.0681	29.6703	2700	0.6934	0.7184	0.8091	0.9267	0.9677	0.8404	0.7619	0.6664	0.9240	0.7284	0.6519	0.5694
0.0518	30.7692	2800	0.7065	0.7183	0.8088	0.9265	0.9692	0.8238	0.7631	0.6789	0.9237	0.7250	0.6514	0.5730
0.0624	31.8681	2900	0.7146	0.7141	0.8005	0.9258	0.9705	0.8284	0.7439	0.6590	0.9230	0.7230	0.6446	0.5658
0.0518	32.9670	3000	0.6993	0.7171	0.8096	0.9261	0.9675	0.8351	0.7605	0.6754	0.9234	0.7252	0.6497	0.5703
0.0658	34.0659	3100	0.7306	0.7162	0.8066	0.9261	0.9686	0.8282	0.7623	0.6674	0.9234	0.7243	0.6502	0.5669
0.0554	35.1648	3200	0.7374	0.7144	0.8021	0.9257	0.9696	0.8256	0.7557	0.6577	0.9229	0.7226	0.6475	0.5645
0.0623	36.2637	3300	0.7468	0.7151	0.8054	0.9256	0.9688	0.8258	0.7566	0.6705	0.9229	0.7220	0.6475	0.5678
0.057	37.3626	3400	0.7597	0.7138	0.8034	0.9253	0.9692	0.8206	0.7584	0.6652	0.9225	0.7207	0.6472	0.5647
0.0644	38.4615	3500	0.7481	0.7146	0.8049	0.9256	0.9681	0.8303	0.7598	0.6614	0.9228	0.7226	0.6482	0.5647
0.0593	39.5604	3600	0.7704	0.7142	0.8040	0.9254	0.9683	0.8289	0.7597	0.6591	0.9227	0.7221	0.6478	0.5640
0.0519	40.6593	3700	0.7748	0.7132	0.8020	0.9254	0.9689	0.8275	0.7565	0.6550	0.9226	0.7218	0.6468	0.5618
0.045	41.7582	3800	0.7814	0.7133	0.8030	0.9252	0.9687	0.8256	0.7577	0.6601	0.9225	0.7212	0.6467	0.5630
0.045	42.8571	3900	0.7831	0.7125	0.8007	0.9252	0.9693	0.8237	0.7553	0.6546	0.9224	0.7202	0.6460	0.5614
0.0523	43.9560	4000	0.7910	0.7125	0.8014	0.9251	0.9690	0.8241	0.7563	0.6564	0.9223	0.7202	0.6460	0.5616
0.0642	45.0549	4100	0.7819	0.7129	0.8019	0.9252	0.9689	0.8253	0.7560	0.6573	0.9224	0.7206	0.6462	0.5623
0.0535	46.1538	4200	0.7885	0.7125	0.8013	0.9251	0.9690	0.8236	0.7570	0.6554	0.9223	0.7202	0.6461	0.5614
0.0474	47.2527	4300	0.7904	0.7123	0.8013	0.9250	0.9690	0.8231	0.7564	0.6566	0.9222	0.7198	0.6457	0.5614
0.0494	48.3516	4400	0.7834	0.7124	0.8015	0.9250	0.9689	0.8239	0.7561	0.6570	0.9223	0.7199	0.6458	0.5615
0.0663	49.4505	4500	0.7882	0.7125	0.8017	0.9250	0.9689	0.8241	0.7558	0.6579	0.9223	0.7200	0.6457	0.5619

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.1+cu130
• Datasets 4.4.1
• Tokenizers 0.22.1