swin-tiny-patch4-window7-224-finetuned-eurosat

This model is a fine-tuned version of microsoft/swin-tiny-patch4-window7-224 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.5712
• Accuracy: 0.8086

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: 5e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.87	5	1.3767	0.5370
1.289	1.91	11	1.3503	0.5494
1.289	2.96	17	1.3712	0.5556
1.0376	4.0	23	1.3064	0.5556
1.0376	4.87	28	1.1062	0.5802
0.8346	5.91	34	0.9249	0.6481
0.7096	6.96	40	0.8947	0.6235
0.7096	8.0	46	0.8626	0.6543
0.6356	8.87	51	0.6820	0.7222
0.6356	9.91	57	0.7249	0.7346
0.5956	10.96	63	0.6818	0.7407
0.5956	12.0	69	0.6111	0.7840
0.5534	12.87	74	0.6026	0.7778
0.519	13.91	80	0.6070	0.7901
0.519	14.96	86	0.5758	0.7963
0.5117	16.0	92	0.5791	0.7840
0.5117	16.87	97	0.5711	0.8025
0.4913	17.39	100	0.5712	0.8086

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2