segformer-b5-finetuned-apple-dms-run7

This model is a fine-tuned version of nvidia/segformer-b5-finetuned-ade-640-640 on the AllanK24/apple-dms-materials dataset. It achieves the following results on the evaluation set:

Loss: 1.3192
Mean Iou: 0.3630
Mean Accuracy: 0.4423
Overall Accuracy: 0.7928
Accuracy Animal Skin: 0.0003
Iou Animal Skin: 0.0003
Accuracy Bone Teeth Horn: 0.0
Iou Bone Teeth Horn: 0.0
Accuracy Brickwork: 0.6738
Iou Brickwork: 0.5345
Accuracy Cardboard: 0.5569
Iou Cardboard: 0.4395
Accuracy Carpet Rug: 0.8605
Iou Carpet Rug: 0.7131
Accuracy Ceiling Tile: 0.8633
Iou Ceiling Tile: 0.7422
Accuracy Ceramic: 0.7179
Iou Ceramic: 0.5672
Accuracy Chalkboard Blackboard: 0.4211
Iou Chalkboard Blackboard: 0.3839
Accuracy Clutter: 0.0
Iou Clutter: 0.0
Accuracy Concrete: 0.5419
Iou Concrete: 0.3820
Accuracy Cork Corkboard: 0.0
Iou Cork Corkboard: 0.0
Accuracy Engineered Stone: 0.0
Iou Engineered Stone: 0.0
Accuracy Fabric Cloth: 0.8896
Iou Fabric Cloth: 0.7605
Accuracy Fiberglass Wool: 0.0
Iou Fiberglass Wool: 0.0
Accuracy Fire: 0.0
Iou Fire: 0.0
Accuracy Foliage: 0.9353
Iou Foliage: 0.8202
Accuracy Food: 0.8894
Iou Food: 0.7479
Accuracy Fur: 0.9129
Iou Fur: 0.7322
Accuracy Gemstone Quartz: 0.0
Iou Gemstone Quartz: 0.0
Accuracy Glass: 0.7364
Iou Glass: 0.5750
Accuracy Hair: 0.8049
Iou Hair: 0.6815
Accuracy Ice: 0.0
Iou Ice: 0.0
Accuracy Leather: 0.4800
Iou Leather: 0.3915
Accuracy Liquid Non-water: 0.0
Iou Liquid Non-water: 0.0
Accuracy Metal: 0.3939
Iou Metal: 0.3004
Accuracy Mirror: 0.5509
Iou Mirror: 0.4350
Accuracy Paint Plaster Enamel: 0.8588
Iou Paint Plaster Enamel: 0.7243
Accuracy Paper: 0.7307
Iou Paper: 0.5634
Accuracy Pearl: 0.0
Iou Pearl: 0.0
Accuracy Photograph Painting: 0.4050
Iou Photograph Painting: 0.3124
Accuracy Plastic Clear: 0.1821
Iou Plastic Clear: 0.1597
Accuracy Plastic Non-clear: 0.4768
Iou Plastic Non-clear: 0.3419
Accuracy Rubber Latex: 0.0
Iou Rubber Latex: 0.0
Accuracy Sand: 0.4155
Iou Sand: 0.3879
Accuracy Skin Lips: 0.8052
Iou Skin Lips: 0.6839
Accuracy Sky: 0.9688
Iou Sky: 0.9233
Accuracy Snow: 0.6109
Iou Snow: 0.5618
Accuracy Soap: 0.0
Iou Soap: 0.0
Accuracy Soil Mud: 0.5931
Iou Soil Mud: 0.3909
Accuracy Sponge: 0.0
Iou Sponge: 0.0
Accuracy Stone Natural: 0.6618
Iou Stone Natural: 0.4872
Accuracy Stone Polished: 0.1331
Iou Stone Polished: 0.1252
Accuracy Styrofoam: 0.0
Iou Styrofoam: 0.0
Accuracy Tile: 0.7938
Iou Tile: 0.6378
Accuracy Wallpaper: 0.5247
Iou Wallpaper: 0.4280
Accuracy Water: 0.9073
Iou Water: 0.7776
Accuracy Wax: 0.0
Iou Wax: 0.0
Accuracy Whiteboard: 0.7118
Iou Whiteboard: 0.5000
Accuracy Wicker: 0.3448
Iou Wicker: 0.3208
Accuracy Wood: 0.8547
Iou Wood: 0.7110
Accuracy Wood Tree: 0.2178
Iou Wood Tree: 0.1922
Accuracy Asphalt: 0.5736
Iou Asphalt: 0.4419

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: 32
eval_batch_size: 16
seed: 42
distributed_type: multi-GPU
num_devices: 8
total_train_batch_size: 256
total_eval_batch_size: 128
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: 0.1
num_epochs: 20
label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Animal Skin	Iou Animal Skin	Accuracy Brickwork	Iou Brickwork	Accuracy Cardboard	Iou Cardboard	Accuracy Carpet Rug	Iou Carpet Rug	Accuracy Ceiling Tile	Iou Ceiling Tile	Accuracy Ceramic	Iou Ceramic	Accuracy Chalkboard Blackboard	Iou Chalkboard Blackboard	Accuracy Concrete	Iou Concrete	Accuracy Fabric Cloth	Iou Fabric Cloth	Accuracy Foliage	Iou Foliage	Accuracy Food	Iou Food	Accuracy Fur	Iou Fur	Accuracy Glass	Iou Glass	Accuracy Hair	Iou Hair	Accuracy Leather	Iou Leather	Accuracy Metal	Iou Metal	Accuracy Mirror	Iou Mirror	Accuracy Paint Plaster Enamel	Iou Paint Plaster Enamel	Accuracy Paper	Iou Paper	Accuracy Photograph Painting	Iou Photograph Painting	Accuracy Plastic Clear	Iou Plastic Clear	Accuracy Plastic Non-clear	Iou Plastic Non-clear	Accuracy Sand	Iou Sand	Accuracy Skin Lips	Iou Skin Lips	Accuracy Sky	Iou Sky	Accuracy Snow	Iou Snow	Accuracy Soil Mud	Iou Soil Mud	Accuracy Stone Natural	Iou Stone Natural	Accuracy Stone Polished	Iou Stone Polished	Accuracy Tile	Iou Tile	Accuracy Wallpaper	Iou Wallpaper	Accuracy Water	Iou Water	Accuracy Whiteboard	Iou Whiteboard	Accuracy Wicker	Iou Wicker	Accuracy Wood	Iou Wood	Accuracy Wood Tree	Iou Wood Tree	Accuracy Asphalt	Iou Asphalt
3.5942	1.7045	150	2.5848	0.0639	0.0960	0.5317	0.0	0.0	0.0	0.0	0.0	0.0	0.0984	0.0952	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.8936	0.4095	0.8037	0.5223	0.0	0.0	0.0	0.0	0.3577	0.2375	0.0	0.0	0.0	0.0	0.0000	0.0000	0.0	0.0	0.7816	0.5720	0.0017	0.0017	0.0	0.0	0.0	0.0	0.0025	0.0025	0.0	0.0	0.0	0.0	0.9472	0.8621	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.3403	0.2667	0.0	0.0	0.0546	0.0544	0.0	0.0	0.0	0.0	0.7127	0.3000	0.0	0.0	0.0	0.0
1.9460	3.4091	300	1.6887	0.1583	0.2074	0.6906	0.0	0.0	0.0	0.0	0.0	0.0	0.7844	0.5407	0.7715	0.6552	0.0	0.0	0.0	0.0	0.0359	0.0351	0.8989	0.6418	0.9376	0.6128	0.6183	0.5662	0.0000	0.0000	0.6617	0.4543	0.4619	0.4115	0.0	0.0	0.0840	0.0743	0.0000	0.0000	0.8458	0.6539	0.6632	0.4467	0.0	0.0	0.0	0.0	0.2981	0.2058	0.0	0.0	0.2557	0.2316	0.9591	0.9030	0.0	0.0	0.0	0.0	0.0602	0.0593	0.0	0.0	0.6902	0.5001	0.0	0.0	0.8989	0.6430	0.0	0.0	0.0	0.0	0.8586	0.5940	0.0	0.0	0.0	0.0
1.5302	5.1136	450	1.4906	0.2292	0.2887	0.7431	0.0	0.0	0.0038	0.0038	0.0003	0.0003	0.8280	0.6343	0.8414	0.7106	0.4694	0.4059	0.0	0.0	0.4276	0.2839	0.8945	0.7071	0.9393	0.7621	0.8639	0.6955	0.7905	0.6364	0.6762	0.5162	0.7001	0.5991	0.0	0.0	0.2660	0.2096	0.4087	0.3370	0.8561	0.6857	0.7286	0.4947	0.1417	0.1271	0.0	0.0	0.3813	0.2647	0.0	0.0	0.6885	0.5667	0.9609	0.9114	0.0	0.0	0.0790	0.0722	0.5492	0.3325	0.0	0.0	0.7704	0.5712	0.0	0.0	0.9052	0.7265	0.0003	0.0003	0.0	0.0	0.8438	0.6633	0.0	0.0	0.0	0.0
1.3887	6.8182	600	1.4086	0.2777	0.3444	0.7652	0.0	0.0	0.5119	0.4067	0.2557	0.2460	0.8398	0.6684	0.8401	0.7281	0.6350	0.5112	0.0	0.0	0.5219	0.3507	0.8760	0.7362	0.9388	0.7927	0.8792	0.7295	0.8919	0.6954	0.6956	0.5350	0.7756	0.6414	0.1100	0.1069	0.3090	0.2431	0.4612	0.3773	0.8686	0.7014	0.7223	0.5287	0.3187	0.2506	0.0000	0.0000	0.3879	0.2852	0.0	0.0	0.7439	0.6247	0.9657	0.9136	0.0	0.0	0.4512	0.2614	0.6270	0.4190	0.0003	0.0003	0.7648	0.6029	0.2160	0.2065	0.9032	0.7619	0.4685	0.3632	0.0	0.0	0.8562	0.6794	0.0	0.0	0.0736	0.0728
1.3155	8.5227	750	1.3680	0.3028	0.3776	0.7759	0.0	0.0	0.6176	0.4674	0.4570	0.3896	0.8520	0.6894	0.8501	0.7356	0.6884	0.5392	0.0001	0.0001	0.5296	0.3721	0.8876	0.7452	0.9376	0.8018	0.8811	0.7457	0.9016	0.7166	0.7120	0.5493	0.7873	0.6602	0.3256	0.2811	0.3465	0.2680	0.5307	0.4008	0.8608	0.7126	0.7335	0.5429	0.3773	0.2864	0.0234	0.0233	0.4252	0.3091	0.0127	0.0127	0.7738	0.6505	0.9669	0.9152	0.0018	0.0018	0.5394	0.2882	0.6169	0.4453	0.0069	0.0069	0.7853	0.6146	0.4590	0.3894	0.9055	0.7681	0.6194	0.3880	0.0	0.0	0.8449	0.6935	0.0010	0.0010	0.3744	0.3335
1.2705	10.2273	900	1.3510	0.3189	0.3975	0.7815	0.0	0.0	0.6430	0.5133	0.5303	0.4237	0.8436	0.6995	0.8634	0.7349	0.7018	0.5473	0.0844	0.0834	0.5602	0.3765	0.8922	0.7486	0.9350	0.8091	0.8901	0.7363	0.9100	0.7149	0.6974	0.5573	0.7983	0.6697	0.3947	0.3269	0.3496	0.2739	0.5309	0.4107	0.8611	0.7171	0.7173	0.5537	0.3948	0.3015	0.0638	0.0620	0.4645	0.3261	0.0730	0.0725	0.7823	0.6639	0.9691	0.9184	0.1254	0.1249	0.6308	0.3122	0.6454	0.4771	0.0419	0.0416	0.7798	0.6248	0.4962	0.4134	0.9030	0.7734	0.6963	0.4199	0.0240	0.0240	0.8508	0.7001	0.0268	0.0266	0.5004	0.4057
1.2421	11.9318	1050	1.3359	0.3373	0.4161	0.7862	0.0	0.0	0.6577	0.5211	0.5421	0.4284	0.8724	0.6988	0.8609	0.7398	0.7091	0.5569	0.1468	0.1418	0.5249	0.3780	0.8887	0.7544	0.9397	0.8091	0.8846	0.7440	0.9056	0.7370	0.7106	0.5658	0.8018	0.6734	0.4367	0.3578	0.3805	0.2904	0.5557	0.4193	0.8567	0.7206	0.7384	0.5554	0.4223	0.3134	0.1021	0.0960	0.4784	0.3344	0.2135	0.2065	0.7943	0.6732	0.9676	0.9195	0.4184	0.4071	0.5614	0.3408	0.6606	0.4714	0.0869	0.0847	0.7862	0.6295	0.5092	0.4181	0.9058	0.7727	0.6929	0.4305	0.1438	0.1415	0.8526	0.7049	0.0807	0.0779	0.5464	0.4240
1.2259	13.6364	1200	1.3284	0.3515	0.4299	0.7896	0.0	0.0	0.6752	0.5282	0.5489	0.4391	0.8585	0.7070	0.8617	0.7427	0.7139	0.5609	0.3240	0.3028	0.5397	0.3802	0.8914	0.7570	0.9389	0.8148	0.8881	0.7454	0.9081	0.7372	0.7239	0.5707	0.8067	0.6762	0.4589	0.3749	0.3816	0.2931	0.5382	0.4271	0.8571	0.7227	0.7244	0.5625	0.4111	0.3133	0.1309	0.1195	0.4867	0.3392	0.3215	0.3059	0.7937	0.6770	0.9692	0.9217	0.5358	0.5029	0.5698	0.3641	0.6698	0.4800	0.1020	0.0986	0.7926	0.6313	0.5189	0.4237	0.9069	0.7749	0.7129	0.4797	0.2450	0.2366	0.8534	0.7084	0.1356	0.1263	0.5620	0.4320
1.2106	15.3409	1350	1.3234	0.3579	0.4372	0.7911	0.0000	0.0000	0.6742	0.5304	0.5591	0.4388	0.8593	0.7095	0.8542	0.7437	0.7155	0.5637	0.3970	0.3624	0.5413	0.3797	0.8883	0.7601	0.9366	0.8182	0.8900	0.7453	0.9128	0.7305	0.7134	0.5700	0.8076	0.6791	0.4678	0.3833	0.3989	0.2984	0.5349	0.4259	0.8618	0.7230	0.7326	0.5615	0.4174	0.3154	0.1600	0.1432	0.4825	0.3411	0.3488	0.3283	0.8042	0.6816	0.9694	0.9215	0.5804	0.5379	0.6104	0.3833	0.6525	0.4851	0.1228	0.1163	0.7942	0.6358	0.5175	0.4257	0.9057	0.7787	0.7096	0.4866	0.3062	0.2894	0.8491	0.7099	0.1876	0.1708	0.5686	0.4375
1.2024	17.0455	1500	1.3213	0.3611	0.4406	0.7922	0.0002	0.0002	0.6728	0.5353	0.5619	0.4388	0.8680	0.7100	0.8600	0.7431	0.7151	0.5654	0.4029	0.3675	0.5378	0.3813	0.8835	0.7619	0.9370	0.8188	0.8889	0.7490	0.9105	0.7367	0.7251	0.5735	0.8064	0.6812	0.4835	0.3907	0.3936	0.2989	0.5487	0.4330	0.8608	0.7236	0.7310	0.5627	0.4099	0.3137	0.1788	0.1573	0.4829	0.3425	0.3858	0.3623	0.8053	0.6833	0.9685	0.9235	0.5946	0.5509	0.5925	0.3832	0.6693	0.4860	0.1320	0.1243	0.7928	0.6372	0.5222	0.4261	0.9063	0.7774	0.7148	0.4967	0.3346	0.3123	0.8549	0.7101	0.2026	0.1818	0.5761	0.4395
1.1973	18.75	1650	1.3192	0.3630	0.4423	0.7928	0.0003	0.0003	0.6738	0.5345	0.5569	0.4395	0.8605	0.7131	0.8633	0.7422	0.7179	0.5672	0.4211	0.3839	0.5419	0.3820	0.8896	0.7605	0.9353	0.8202	0.8894	0.7479	0.9129	0.7322	0.7364	0.5750	0.8049	0.6815	0.4800	0.3915	0.3939	0.3004	0.5509	0.4350	0.8588	0.7243	0.7307	0.5634	0.4050	0.3124	0.1821	0.1597	0.4768	0.3419	0.4155	0.3879	0.8052	0.6839	0.9688	0.9233	0.6109	0.5618	0.5931	0.3909	0.6618	0.4872	0.1331	0.1252	0.7938	0.6378	0.5247	0.4280	0.9073	0.7776	0.7118	0.5000	0.3448	0.3208	0.8547	0.7110	0.2178	0.1922	0.5736	0.4419

Framework versions

Transformers 5.0.0
Pytorch 2.9.1+cu128
Datasets 4.5.0
Tokenizers 0.22.2

Downloads last month: 10

Safetensors

Model size

84.6M params

Tensor type

F32

Model tree for AllanK24/segformer-b5-finetuned-apple-dms-run7

Base model

nvidia/segformer-b5-finetuned-ade-640-640

Finetuned

(22)

this model