Aerial Drone Image Segmentation

🚀 Aerial-Drone-Image-Segmentation

This model is designed for aerial drone image segmentation, leveraging a fine - tuned version of [nvidia/mit - b0](https://huggingface.co/nvidia/mit - b0) to achieve high - quality segmentation results.

🚀 Quick Start

This section provides a high - level overview of the model and its performance on the evaluation set.

• Loss: 0.8852
• Mean Iou: 0.2994
• Mean Accuracy: 0.3923
• Overall Accuracy: 0.7774

📚 Documentation

Model description

This model is a fine - tuned version of [nvidia/mit - b0](https://huggingface.co/nvidia/mit - b0).

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e - 05
• train_batch_size: 24
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
• lr_scheduler_type: linear
• num_epochs: 50

Evaluation Results

{'mean_iou': 0.27989828118195953,
 'mean_accuracy': 0.3712316062110093,
 'overall_accuracy': 0.7671712239583334,
 'per_category_iou': array([       nan, 0.8560476 , 0.32234631, 0.76880948, 0.57517691,
        0.43877125, 0.00114888, 0.14091442, 0.51807365, 0.76964765,
        0.27391949, 0.        , 0.        , 0.        , 0.        ,
        0.05778175, 0.        , 0.45566807, 0.        , 0.25864545,
        0.48767764, 0.        , 0.23313364,        nan]),
 'per_category_accuracy': array([       nan, 0.96170675, 0.43993514, 0.86977593, 0.8149788 ,
        0.49739671, 0.00114987, 0.14445379, 0.80978302, 0.88661108,
        0.46787116, 0.        , 0.        , 0.        , 0.        ,
        0.05947339, 0.        , 0.55639324, 0.        , 0.38358184,
        0.761303  , 0.        , 0.51268161,        nan])}

Training results


Framework versions

• Transformers 4.38.1
• Pytorch 2.1.2
• Datasets 2.1.0
• Tokenizers 0.15.2

📄 License

The model uses an "other" license.