upernet-swin-small-finetuned

This model is a fine-tuned version of openmmlab/upernet-swin-small on the jpodivin/plantorgans dataset. It achieves the following results on the evaluation set:

• Loss: 0.2914
• Mean Iou: 0.4182
• Mean Accuracy: 0.5282
• Overall Accuracy: 0.7341
• Accuracy Void: nan
• Accuracy Fruit: 0.8590
• Accuracy Leaf: 0.7032
• Accuracy Flower: 0.0
• Accuracy Stem: 0.5505
• Iou Void: 0.0
• Iou Fruit: 0.8554
• Iou Leaf: 0.6976
• Iou Flower: 0.0
• Iou Stem: 0.5381
• Median Iou: 0.5381

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: 0.0006
• train_batch_size: 10
• eval_batch_size: 10
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Void	Accuracy Fruit	Accuracy Leaf	Accuracy Flower	Accuracy Stem	Iou Void	Iou Fruit	Iou Leaf	Iou Flower	Iou Stem	Median Iou
0.8566	1.0	575	0.3365	0.3723	0.4705	0.6560	nan	0.8000	0.6122	0.0	0.4699	0.0	0.7976	0.6041	0.0	0.4598	0.4598
0.3338	2.0	1150	0.3030	0.3922	0.4937	0.7155	nan	0.8558	0.7024	0.0	0.4166	0.0	0.8517	0.6972	0.0	0.4119	0.4119
0.3477	3.0	1725	0.2914	0.4182	0.5282	0.7341	nan	0.8590	0.7032	0.0	0.5505	0.0	0.8554	0.6976	0.0	0.5381	0.5381

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0