segformerSAAD2

Saad287/segformerSAAD2

segformerSAAD2 is a image segmentation model from Saad287. Use it for the image segmentation task on the model card, and read the license before you ship it in a product. The card lists the license as other.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

7% of all-time downloads

All-time downloads

Public

Parameters

3.7M

15 MB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors14.9 MB · 99%

At a glance

Task: Image Segmentation
License: other
Model type: segformer
Access: Public
Created: Aug 22, 2024
Updated: Aug 22, 2024
SHA: dbde36ae

Base models

nvidia/mit-b0

Model card

From the Hugging Face model README

segformerSAAD2

This model is a fine-tuned version of nvidia/mit-b0 on the Saad287/SixGUN dataset. It achieves the following results on the evaluation set:

Loss: 0.0658
Mean Iou: 0.8885
Mean Accuracy: 0.9302
Overall Accuracy: 0.9934
Accuracy Bkg: 0.9977
Accuracy Knife: 0.8700
Accuracy Gun: 0.9228
Iou Bkg: 0.9940
Iou Knife: 0.8345
Iou Gun: 0.8370

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: 23
eval_batch_size: 23
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 500

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Bkg	Accuracy Knife	Accuracy Gun	Iou Bkg	Iou Knife	Iou Gun
0.9085	10.0	10	1.0050	0.5155	0.9279	0.9160	0.9152	0.9126	0.9559	0.9144	0.3135	0.3185
0.7013	20.0	20	0.8010	0.6030	0.8080	0.9590	0.9692	0.6715	0.7832	0.9590	0.4442	0.4060
0.5991	30.0	30	0.6384	0.6286	0.7939	0.9648	0.9765	0.6946	0.7107	0.9647	0.5156	0.4054
0.5105	40.0	40	0.5393	0.6328	0.8246	0.9632	0.9727	0.7618	0.7393	0.9631	0.5383	0.3970
0.4642	50.0	50	0.4625	0.6478	0.8186	0.9668	0.9769	0.7866	0.6922	0.9666	0.5662	0.4108
0.4301	60.0	60	0.4175	0.6580	0.8092	0.9696	0.9805	0.7819	0.6650	0.9693	0.5757	0.4291
0.3623	70.0	70	0.3713	0.7008	0.8249	0.9758	0.9861	0.7827	0.7061	0.9757	0.6257	0.5010
0.3356	80.0	80	0.3247	0.7648	0.8740	0.9825	0.9899	0.8066	0.8254	0.9827	0.6695	0.6422
0.2975	90.0	90	0.2862	0.7936	0.8754	0.9857	0.9932	0.8001	0.8330	0.9860	0.6986	0.6962
0.2722	100.0	100	0.2656	0.8099	0.8962	0.9869	0.9931	0.8202	0.8753	0.9873	0.7126	0.7296
0.2465	110.0	110	0.2377	0.8192	0.8904	0.9880	0.9946	0.8170	0.8597	0.9884	0.7251	0.7440
0.2221	120.0	120	0.2163	0.8311	0.9018	0.9889	0.9949	0.8265	0.8840	0.9895	0.7391	0.7648
0.2082	130.0	130	0.2007	0.8341	0.8971	0.9892	0.9955	0.8297	0.8662	0.9896	0.7500	0.7628
0.1974	140.0	140	0.1928	0.8462	0.9026	0.9901	0.9961	0.8390	0.8729	0.9906	0.7685	0.7795
0.1823	150.0	150	0.1720	0.8464	0.8974	0.9903	0.9966	0.8324	0.8632	0.9907	0.7710	0.7775
0.1743	160.0	160	0.1598	0.8533	0.9014	0.9908	0.9969	0.8307	0.8766	0.9913	0.7799	0.7888
0.1577	170.0	170	0.1560	0.8592	0.9150	0.9911	0.9963	0.8514	0.8973	0.9917	0.7892	0.7967
0.1468	180.0	180	0.1432	0.8632	0.9079	0.9915	0.9972	0.8406	0.8858	0.9920	0.7964	0.8012
0.1397	190.0	190	0.1323	0.8641	0.9128	0.9915	0.9969	0.8503	0.8911	0.9920	0.7993	0.8009
0.1305	200.0	200	0.1264	0.8659	0.9102	0.9917	0.9972	0.8468	0.8867	0.9922	0.8024	0.8032
0.1228	210.0	210	0.1209	0.8730	0.9188	0.9922	0.9972	0.8552	0.9038	0.9927	0.8119	0.8144
0.1179	220.0	220	0.1123	0.8730	0.9172	0.9922	0.9973	0.8527	0.9016	0.9928	0.8128	0.8133
0.1126	230.0	230	0.1078	0.8742	0.9203	0.9923	0.9972	0.8568	0.9069	0.9929	0.8148	0.8149
0.1066	240.0	240	0.1029	0.8758	0.9216	0.9924	0.9972	0.8587	0.9090	0.9930	0.8176	0.8169
0.1026	250.0	250	0.0978	0.8786	0.9224	0.9926	0.9974	0.8648	0.9050	0.9931	0.8230	0.8197
0.1012	260.0	260	0.0963	0.8797	0.9271	0.9927	0.9972	0.8659	0.9182	0.9933	0.8242	0.8217
0.0952	270.0	270	0.0926	0.8800	0.9229	0.9927	0.9975	0.8607	0.9104	0.9933	0.8250	0.8218
0.0938	280.0	280	0.0895	0.8819	0.9238	0.9929	0.9975	0.8649	0.9090	0.9934	0.8284	0.8237
0.0893	290.0	290	0.0860	0.8832	0.9260	0.9930	0.9975	0.8686	0.9120	0.9935	0.8301	0.8261
0.0868	300.0	300	0.0835	0.8831	0.9243	0.9930	0.9976	0.8646	0.9106	0.9935	0.8301	0.8257
0.0853	310.0	310	0.0820	0.8834	0.9257	0.9930	0.9975	0.8664	0.9132	0.9936	0.8307	0.8260
0.0817	320.0	320	0.0786	0.8867	0.9291	0.9932	0.9976	0.8707	0.9190	0.9938	0.8343	0.8319
0.0823	330.0	330	0.0785	0.8851	0.9261	0.9931	0.9977	0.8677	0.9129	0.9937	0.8326	0.8291
0.0858	340.0	340	0.0768	0.8860	0.9272	0.9932	0.9976	0.8680	0.9159	0.9938	0.8330	0.8312
0.078	350.0	350	0.0744	0.8853	0.9253	0.9931	0.9977	0.8648	0.9133	0.9937	0.8315	0.8308
0.0755	360.0	360	0.0739	0.8872	0.9290	0.9933	0.9976	0.8694	0.9201	0.9939	0.8337	0.8342
0.0752	370.0	370	0.0727	0.8880	0.9307	0.9933	0.9976	0.8721	0.9224	0.9939	0.8351	0.8349
0.0727	380.0	380	0.0719	0.8858	0.9269	0.9932	0.9977	0.8665	0.9166	0.9938	0.8318	0.8318
0.0725	390.0	390	0.0709	0.8873	0.9298	0.9933	0.9976	0.8705	0.9215	0.9939	0.8347	0.8333
0.0725	400.0	400	0.0695	0.8870	0.9283	0.9933	0.9977	0.8686	0.9188	0.9939	0.8331	0.8340
0.0718	410.0	410	0.0689	0.8890	0.9309	0.9934	0.9976	0.8706	0.9244	0.9940	0.8351	0.8378
0.0695	420.0	420	0.0679	0.8872	0.9284	0.9933	0.9977	0.8689	0.9186	0.9939	0.8335	0.8341
0.0707	430.0	430	0.0676	0.8881	0.9307	0.9933	0.9976	0.8686	0.9258	0.9940	0.8334	0.8370
0.0684	440.0	440	0.0668	0.8883	0.9298	0.9934	0.9977	0.8693	0.9225	0.9940	0.8339	0.8372
0.0692	450.0	450	0.0668	0.8887	0.9310	0.9934	0.9976	0.8708	0.9245	0.9940	0.8350	0.8372
0.0681	460.0	460	0.0661	0.8889	0.9296	0.9934	0.9977	0.8696	0.9216	0.9940	0.8344	0.8382
0.0688	470.0	470	0.0661	0.8888	0.9305	0.9934	0.9977	0.8708	0.9230	0.9940	0.8352	0.8372
0.0675	480.0	480	0.0659	0.8888	0.9307	0.9934	0.9976	0.8717	0.9228	0.9940	0.8353	0.8370
0.0679	490.0	490	0.0659	0.8886	0.9300	0.9934	0.9977	0.8708	0.9214	0.9940	0.8352	0.8367
0.0668	500.0	500	0.0658	0.8885	0.9302	0.9934	0.9977	0.8700	0.9228	0.9940	0.8345	0.8370

Framework versions

Transformers 4.42.4
Pytorch 2.3.1+cu121
Datasets 2.21.0
Tokenizers 0.19.1

Task: Image Segmentation
Type: segformer
License: other
Created: Aug 22, 2024
Updated: Aug 22, 2024