SegFormer_Clean_Set1_Grayscale_mit-b5_Grayscale

This model is a fine-tuned version of nvidia/mit-b5 on the Hasano20/Clean_Set1_Grayscale dataset. It achieves the following results on the evaluation set:

• Loss: 0.0178
• Mean Iou: 0.9760
• Mean Accuracy: 0.9847
• Overall Accuracy: 0.9949
• Accuracy Background: 0.9976
• Accuracy Melt: 0.9586
• Accuracy Substrate: 0.9978
• Iou Background: 0.9959
• Iou Melt: 0.9408
• Iou Substrate: 0.9912

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: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 50

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
0.0882	5.5556	50	0.1733	0.7543	0.7985	0.9397	0.9594	0.4375	0.9986	0.9562	0.4189	0.8880
0.0295	11.1111	100	0.0270	0.9580	0.9736	0.9907	0.9965	0.9302	0.9940	0.9918	0.8978	0.9843
0.0143	16.6667	150	0.0260	0.9561	0.9798	0.9901	0.9970	0.9541	0.9884	0.9910	0.8938	0.9836
0.0095	22.2222	200	0.0224	0.9645	0.9747	0.9926	0.9985	0.9293	0.9962	0.9944	0.9119	0.9872
0.0083	27.7778	250	0.0180	0.9742	0.9819	0.9945	0.9982	0.9498	0.9977	0.9955	0.9366	0.9905
0.0072	33.3333	300	0.0175	0.9751	0.9838	0.9947	0.9984	0.9563	0.9968	0.9957	0.9388	0.9909
0.0073	38.8889	350	0.0177	0.9758	0.9854	0.9948	0.9970	0.9613	0.9978	0.9957	0.9406	0.9912
0.0054	44.4444	400	0.0179	0.9758	0.9844	0.9949	0.9978	0.9579	0.9976	0.9959	0.9404	0.9911
0.0052	50.0	450	0.0178	0.9760	0.9847	0.9949	0.9976	0.9586	0.9978	0.9959	0.9408	0.9912

Framework versions

• Transformers 4.41.2
• Pytorch 2.0.1+cu117
• Datasets 2.19.2
• Tokenizers 0.19.1