SegFormer_mit-b5_Clean-Set3-Grayscale_Augmented_Medium_16

Hasano20/SegFormer_mit-b5_Clean-Set3-Grayscale_Augmented_Medium_16

SegFormer_mit-b5_Clean-Set3-Grayscale_Augmented_Medium_16 is a machine learning model from Hasano20. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as other.

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

Downloads · 30 days

23% of all-time downloads

All-time downloads

Public

Parameters

84.6M

339 MB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors339 MB · 100%

At a glance

Library: transformers
License: other
Model type: segformer
Access: Public
Created: Jul 22, 2024
Updated: Jul 22, 2024
SHA: bd486770

Base models

nvidia/mit-b5

Model card

From the Hugging Face model README

SegFormer_mit-b5_Clean-Set3-Grayscale_Augmented_Medium_16

This model is a fine-tuned version of nvidia/mit-b5 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0121
Mean Iou: 0.9823
Mean Accuracy: 0.9920
Overall Accuracy: 0.9954
Accuracy Background: 0.9974
Accuracy Melt: 0.9828
Accuracy Substrate: 0.9958
Iou Background: 0.9943
Iou Melt: 0.9594
Iou Substrate: 0.9932

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: 0.0002
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Melt	Accuracy Substrate	Iou Background	Iou Melt	Iou Substrate
0.1196	0.7937	50	0.1076	0.8582	0.8965	0.9626	0.9674	0.7265	0.9955	0.9625	0.6696	0.9424
0.2728	1.5873	100	0.0878	0.8762	0.9239	0.9665	0.9622	0.8161	0.9935	0.9611	0.7150	0.9525
0.2668	2.3810	150	0.1131	0.8710	0.9238	0.9639	0.9971	0.8140	0.9602	0.9620	0.7076	0.9432
0.0337	3.1746	200	0.0610	0.9173	0.9613	0.9778	0.9709	0.9208	0.9923	0.9685	0.8110	0.9723
0.0443	3.9683	250	0.0295	0.9527	0.9665	0.9885	0.9924	0.9095	0.9977	0.9902	0.8867	0.9812
0.0283	4.7619	300	0.0220	0.9652	0.9781	0.9915	0.9965	0.9429	0.9950	0.9910	0.9175	0.9871
0.0166	5.5556	350	0.0193	0.9683	0.9837	0.9922	0.9972	0.9609	0.9929	0.9925	0.9249	0.9876
0.0218	6.3492	400	0.0190	0.9691	0.9871	0.9922	0.9975	0.9730	0.9909	0.9919	0.9277	0.9879
0.0178	7.1429	450	0.0157	0.9752	0.9853	0.9938	0.9981	0.9626	0.9951	0.9925	0.9424	0.9909
0.0165	7.9365	500	0.0151	0.9771	0.9878	0.9941	0.9966	0.9711	0.9957	0.9931	0.9470	0.9911
0.0136	8.7302	550	0.0137	0.9785	0.9902	0.9945	0.9955	0.9792	0.9959	0.9930	0.9508	0.9918
0.0127	9.5238	600	0.0128	0.9798	0.9896	0.9948	0.9977	0.9758	0.9955	0.9937	0.9536	0.9923
0.0117	10.3175	650	0.0123	0.9809	0.9895	0.9951	0.9974	0.9747	0.9964	0.9939	0.9561	0.9927
0.011	11.1111	700	0.0125	0.9805	0.9923	0.9950	0.9974	0.9848	0.9946	0.9938	0.9552	0.9925
0.0108	11.9048	750	0.0123	0.9809	0.9915	0.9951	0.9975	0.9818	0.9952	0.9940	0.9561	0.9926
0.0135	12.6984	800	0.0126	0.9808	0.9920	0.9950	0.9979	0.9834	0.9946	0.9941	0.9558	0.9924
0.0089	13.4921	850	0.0123	0.9814	0.9923	0.9952	0.9968	0.9844	0.9957	0.9940	0.9574	0.9929
0.0077	14.2857	900	0.0119	0.9819	0.9911	0.9953	0.9976	0.9797	0.9959	0.9942	0.9586	0.9930
0.0069	15.0794	950	0.0122	0.9822	0.9914	0.9954	0.9973	0.9807	0.9961	0.9943	0.9591	0.9931
0.0069	15.8730	1000	0.0120	0.9822	0.9920	0.9954	0.9975	0.9828	0.9957	0.9944	0.9592	0.9931
0.0089	16.6667	1050	0.0120	0.9824	0.9914	0.9955	0.9976	0.9807	0.9961	0.9943	0.9595	0.9932
0.0072	17.4603	1100	0.0121	0.9823	0.9920	0.9954	0.9974	0.9828	0.9958	0.9943	0.9594	0.9932

Framework versions

Transformers 4.41.2
Pytorch 2.0.1+cu117
Datasets 2.19.2
Tokenizers 0.19.1

Library: transformers
Type: segformer
License: other
Created: Jul 22, 2024
Updated: Jul 22, 2024