dushj98/aerial-resnet50-real-only-aug-plateaulr-ft

resnet50 trained on dushj98/aerial_real_only

This model is a fine-tuned version of resnet50 on the dushj98/aerial_real_only dataset.

Model Details

• Model Type: resnet50
• Training Dataset: dushj98/aerial_real_only
• Number of classes: 13

Training Metadata

• model: resnet50
• epochs: 25
• lr: 0.0003
• batch_size: 32
• dataset: dushj98/aerial_real_only
• random_seed: 42
• feature_extraction: False
• training_mode: ft
• data_augmentation: True
• num_classes: 13
• patience: 10
• scheduler: plateau
• best_val_acc: 0.5020727775218793
• best_val_loss: 2.9368316604485987
• best_epoch: 20
• val_acc_history: [0.44587747581759557, 0.47005988023952094, 0.45163519115614925, 0.44173192077383694, 0.44587747581759557, 0.45071395670198067, 0.4815753109166283, 0.49147858129894056, 0.4944725932749885, 0.482726853984339, 0.4834177798249655, 0.48595117457392906, 0.48433901427913406, 0.49101796407185627, 0.49216950713956703, 0.5002303086135421, 0.4944725932749885, 0.49884845693228924, 0.49585444495624137, 0.501612160294795, 0.5020727775218793, 0.4926301243666513, 0.4960847535697835, 0.49332105020727773, 0.5009212344541686]
• train_time: 895.287643564865
• lr_history: [0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.00015, 0.00015, 0.00015, 0.00015, 7.5e-05, 7.5e-05, 7.5e-05, 7.5e-05, 3.75e-05, 3.75e-05, 3.75e-05, 3.75e-05, 1.875e-05, 1.875e-05, 1.875e-05, 1.875e-05, 9.375e-06, 9.375e-06, 9.375e-06]