mask2former-swin-base-coco-instance-cvppp-a1-ft

This model is a fine-tuned version of facebook/mask2former-swin-base-coco-instance on the fengchen025/cvppp-a1 dataset. It achieves the following results on the evaluation set:

• Loss: 8.1314
• Map: 0.4892
• Map 50: 0.6847
• Map 75: 0.5666
• Map Small: 0.4391
• Map Medium: 0.5228
• Map Large: -1.0
• Mar 1: 0.0532
• Mar 10: 0.4909
• Mar 100: 0.7390
• Mar Small: 0.5736
• Mar Medium: 0.8257
• Mar Large: -1.0
• Map Per Class: 0.4892
• Mar 100 Per Class: 0.7390

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 50.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Map	Map 50	Map 75	Map Small	Map Medium	Map Large	Mar 1	Mar 10	Mar 100	Mar Small	Mar Medium	Mar Large	Map Per Class	Mar 100 Per Class
39.3962	1.0	15	21.0739	0.2211	0.4731	0.1821	0.1131	0.2731	-1.0	0.0286	0.2740	0.5117	0.2849	0.6307	-1.0	0.2211	0.5117
19.3166	2.0	30	16.2822	0.3542	0.5935	0.3793	0.2599	0.4156	-1.0	0.0422	0.3883	0.5760	0.3792	0.6792	-1.0	0.3542	0.5760
16.0327	3.0	45	14.8551	0.3826	0.6235	0.3937	0.2844	0.4328	-1.0	0.0448	0.4208	0.6058	0.4226	0.7020	-1.0	0.3826	0.6058
14.9212	4.0	60	14.1509	0.3870	0.6213	0.4268	0.3053	0.4422	-1.0	0.0442	0.4136	0.5981	0.4094	0.6970	-1.0	0.3870	0.5981
14.1777	5.0	75	13.3047	0.3968	0.6533	0.4284	0.3201	0.4410	-1.0	0.0435	0.4156	0.6312	0.4453	0.7287	-1.0	0.3968	0.6312
13.3037	6.0	90	12.1355	0.4169	0.6756	0.4283	0.3320	0.4631	-1.0	0.0455	0.4422	0.6448	0.4736	0.7347	-1.0	0.4169	0.6448
12.7922	7.0	105	11.9884	0.4204	0.6434	0.4713	0.3444	0.4678	-1.0	0.0448	0.4506	0.6669	0.5038	0.7525	-1.0	0.4204	0.6669
11.8428	8.0	120	11.1682	0.4221	0.6415	0.4704	0.3662	0.4629	-1.0	0.0455	0.4532	0.6701	0.5189	0.7495	-1.0	0.4221	0.6701
11.4887	9.0	135	11.3821	0.4237	0.6425	0.4809	0.3503	0.4697	-1.0	0.0455	0.4571	0.6675	0.5019	0.7545	-1.0	0.4237	0.6675
11.4137	10.0	150	10.7176	0.4526	0.6702	0.5098	0.3592	0.5070	-1.0	0.0565	0.4610	0.6870	0.5245	0.7723	-1.0	0.4526	0.6870
11.0166	11.0	165	10.6001	0.4586	0.6851	0.5253	0.3706	0.5087	-1.0	0.0500	0.4571	0.6838	0.5472	0.7554	-1.0	0.4586	0.6838
10.7357	12.0	180	10.4470	0.4478	0.6602	0.5013	0.3750	0.4918	-1.0	0.0519	0.4688	0.6955	0.5264	0.7842	-1.0	0.4478	0.6955
10.6097	13.0	195	10.4092	0.4738	0.6906	0.5290	0.3608	0.5398	-1.0	0.0565	0.4695	0.6870	0.5245	0.7723	-1.0	0.4738	0.6870
10.1015	14.0	210	9.9768	0.4372	0.6690	0.4708	0.3689	0.4776	-1.0	0.0519	0.4597	0.6727	0.5113	0.7574	-1.0	0.4372	0.6727
10.0756	15.0	225	10.3155	0.4536	0.6756	0.4957	0.3664	0.5020	-1.0	0.0455	0.4649	0.7052	0.5453	0.7891	-1.0	0.4536	0.7052
9.773	16.0	240	10.4804	0.4510	0.6772	0.5103	0.3897	0.4925	-1.0	0.0448	0.4617	0.7104	0.5547	0.7921	-1.0	0.4510	0.7104
9.9397	17.0	255	9.5911	0.4657	0.6811	0.5238	0.4039	0.5043	-1.0	0.0526	0.4766	0.7058	0.5453	0.7901	-1.0	0.4657	0.7058
9.4745	18.0	270	9.4730	0.4550	0.6912	0.4941	0.3794	0.5014	-1.0	0.0519	0.4695	0.7058	0.5358	0.7950	-1.0	0.4550	0.7058
9.4424	19.0	285	9.5308	0.4389	0.6686	0.4737	0.3505	0.4841	-1.0	0.0519	0.4701	0.6825	0.4981	0.7792	-1.0	0.4389	0.6825
9.2619	20.0	300	10.2104	0.4552	0.6874	0.5007	0.3839	0.5102	-1.0	0.0429	0.4669	0.6838	0.5283	0.7653	-1.0	0.4552	0.6838
9.2045	21.0	315	9.6575	0.4618	0.6738	0.5172	0.3904	0.5113	-1.0	0.0591	0.4630	0.7006	0.5321	0.7891	-1.0	0.4618	0.7006
9.2804	22.0	330	9.4810	0.4593	0.6918	0.5111	0.3870	0.5076	-1.0	0.0578	0.4649	0.7110	0.5472	0.7970	-1.0	0.4593	0.7110
9.0051	23.0	345	9.6407	0.4498	0.6726	0.4965	0.3774	0.5037	-1.0	0.0532	0.4669	0.6994	0.5170	0.7950	-1.0	0.4498	0.6994
9.0604	24.0	360	9.4909	0.4700	0.6807	0.5087	0.4251	0.5149	-1.0	0.0578	0.4727	0.7039	0.5302	0.7950	-1.0	0.4700	0.7039
8.8417	25.0	375	9.5084	0.4638	0.6895	0.5165	0.4111	0.5066	-1.0	0.0461	0.4662	0.7065	0.5396	0.7941	-1.0	0.4638	0.7065
8.6481	26.0	390	9.2261	0.4792	0.6964	0.5357	0.3961	0.5348	-1.0	0.0584	0.4812	0.7195	0.5698	0.7980	-1.0	0.4792	0.7195
8.8093	27.0	405	9.0720	0.4771	0.6920	0.5341	0.4281	0.5160	-1.0	0.0513	0.4682	0.7175	0.5698	0.7950	-1.0	0.4771	0.7175
8.4759	28.0	420	8.7442	0.4779	0.7007	0.5180	0.4187	0.5212	-1.0	0.0571	0.4688	0.7214	0.5528	0.8099	-1.0	0.4779	0.7214
8.1716	29.0	435	8.7102	0.4749	0.6801	0.5537	0.4114	0.5306	-1.0	0.0532	0.4727	0.7149	0.5377	0.8079	-1.0	0.4749	0.7149
7.9733	30.0	450	8.6251	0.4845	0.6957	0.5308	0.4179	0.5290	-1.0	0.0591	0.4877	0.7195	0.5528	0.8069	-1.0	0.4845	0.7195
7.9114	31.0	465	9.0874	0.4729	0.6975	0.5221	0.4137	0.5087	-1.0	0.0487	0.4786	0.7156	0.5774	0.7881	-1.0	0.4729	0.7156
7.9404	32.0	480	8.4683	0.4768	0.6950	0.5331	0.4201	0.5192	-1.0	0.0591	0.4799	0.7234	0.5698	0.8040	-1.0	0.4768	0.7234
8.0908	33.0	495	9.1853	0.4768	0.6974	0.5357	0.3872	0.5308	-1.0	0.0532	0.4857	0.7143	0.5396	0.8059	-1.0	0.4768	0.7143
7.9822	34.0	510	8.5842	0.4712	0.6871	0.5407	0.4397	0.5086	-1.0	0.0591	0.4714	0.7247	0.5585	0.8119	-1.0	0.4712	0.7247
7.9252	35.0	525	8.9616	0.4670	0.6817	0.5517	0.4207	0.5048	-1.0	0.0519	0.4734	0.7208	0.5642	0.8030	-1.0	0.4670	0.7208
7.7363	36.0	540	8.7401	0.4718	0.6898	0.5241	0.4015	0.5241	-1.0	0.0526	0.4877	0.7221	0.5415	0.8168	-1.0	0.4718	0.7221
7.6896	37.0	555	8.8058	0.4806	0.6912	0.5278	0.4108	0.5280	-1.0	0.0591	0.4799	0.7305	0.5547	0.8228	-1.0	0.4806	0.7305
7.6104	38.0	570	8.1799	0.4748	0.6876	0.5268	0.4338	0.5144	-1.0	0.0532	0.4825	0.7247	0.5547	0.8139	-1.0	0.4748	0.7247
7.2744	39.0	585	8.4149	0.4737	0.6780	0.5461	0.3985	0.5228	-1.0	0.0597	0.4779	0.7201	0.5415	0.8139	-1.0	0.4737	0.7201
7.2528	40.0	600	8.5222	0.4797	0.6828	0.5380	0.4185	0.5213	-1.0	0.0526	0.4831	0.7234	0.5566	0.8109	-1.0	0.4797	0.7234
7.0455	41.0	615	8.4200	0.4782	0.6931	0.5453	0.4118	0.5284	-1.0	0.0584	0.4818	0.7247	0.5547	0.8139	-1.0	0.4782	0.7247
7.0402	42.0	630	8.1926	0.4898	0.6884	0.5483	0.4347	0.5366	-1.0	0.0591	0.4890	0.7344	0.5660	0.8228	-1.0	0.4898	0.7344
7.0456	43.0	645	8.5564	0.4859	0.6750	0.5503	0.4107	0.5366	-1.0	0.0610	0.4955	0.7383	0.5604	0.8317	-1.0	0.4859	0.7383
6.9132	44.0	660	8.7082	0.4910	0.6897	0.5585	0.4342	0.5394	-1.0	0.0532	0.4987	0.7357	0.5585	0.8287	-1.0	0.4910	0.7357
7.1349	45.0	675	8.1748	0.4908	0.6962	0.5585	0.4296	0.5403	-1.0	0.0526	0.4896	0.7364	0.5792	0.8188	-1.0	0.4908	0.7364
6.9486	46.0	690	8.2447	0.4772	0.6904	0.5290	0.4158	0.5266	-1.0	0.0597	0.4916	0.7338	0.5547	0.8277	-1.0	0.4772	0.7338
7.1528	47.0	705	8.2414	0.4902	0.6889	0.5501	0.4370	0.5336	-1.0	0.0526	0.4870	0.7344	0.5736	0.8188	-1.0	0.4902	0.7344
7.0135	48.0	720	7.6569	0.4972	0.7006	0.5627	0.4546	0.5327	-1.0	0.0604	0.4981	0.7468	0.5906	0.8287	-1.0	0.4972	0.7468
6.6802	49.0	735	7.9771	0.4954	0.6861	0.5512	0.4406	0.5370	-1.0	0.0539	0.5	0.7409	0.5736	0.8287	-1.0	0.4954	0.7409
6.7759	50.0	750	8.1314	0.4892	0.6847	0.5666	0.4391	0.5228	-1.0	0.0532	0.4909	0.7390	0.5736	0.8257	-1.0	0.4892	0.7390

Framework versions

• Transformers 4.44.0
• Pytorch 2.1.0.dev20230618
• Datasets 2.11.0
• Tokenizers 0.19.1