Modelpublic

Thibaut/route_background_semantic

sourceHugging Faceotherupdated 2y agoView on Hugging Face

0likes8downloads

Model Card

routebackgroundsemantic

This model is a fine-tuned version of nvidia/segformer-b3-finetuned-cityscapes-1024-1024 on the Logiroad/routebackgroundsemantic dataset. It achieves the following results on the evaluation set:

—Loss: 0.2360
—Mean Iou: 0.1916
—Mean Accuracy: 0.2447
—Overall Accuracy: 0.2962
—Accuracy Unlabeled: nan
—Accuracy Découpe: 0.2865
—Accuracy Reflet météo: 0.0
—Accuracy Autre réparation: 0.3437
—Accuracy Glaçage ou ressuage: 0.0386
—Accuracy Emergence: 0.5549
—Iou Unlabeled: 0.0
—Iou Découpe: 0.2515
—Iou Reflet météo: 0.0
—Iou Autre réparation: 0.3230
—Iou Glaçage ou ressuage: 0.0369
—Iou Emergence: 0.5379

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
—trainbatchsize: 4
—evalbatchsize: 4
—seed: 1337
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: polynomial
—training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Découpe	Accuracy Autre réparation	Accuracy Glaçage ou ressuage	Accuracy Emergence	Iou Découpe	Iou Autre réparation	Iou Glaçage ou ressuage	Iou Emergence
0.2715	1.0	2427	0.2682	0.0521	0.0669	0.1828	nan	0.0813	0.2533	0.0	0.0	0.0766	0.2362	0.0	0.0
0.2815	2.0	4854	0.2682	0.1165	0.1436	0.1593	nan	0.1108	0.1982	0.0	0.4090	0.1014	0.1916	0.0	0.4057
0.2638	3.0	7281	0.2420	0.1664	0.2100	0.2564	nan	0.2346	0.3039	0.0030	0.5085	0.2128	0.2854	0.0030	0.4973
0.2703	4.0	9708	0.2333	0.1941	0.2475	0.3074	nan	0.2843	0.3612	0.0446	0.5473	0.2512	0.3383	0.0429	0.5320
0.2197	4.1203	10000	0.2360	0.1916	0.2447	0.2962	nan	0.2865	0.3437	0.0386	0.5549	0.2515	0.3230	0.0369	0.5379

Framework versions

—Transformers 4.46.1
—Pytorch 2.3.0
—Datasets 3.1.0
—Tokenizers 0.20.3