- Information Gathering: Gathered images using Google Earth.
- Image Collection: Collected images for the project's dataset.
- Image Preprocessing: Prepared and processed images for further use.
- Annotation: Annotated the collected images for training the model.
- Environment Setup: Configured the environment for model training.
- Data Training: Divided collected data into 80% training and 20% testing datasets. Ran 60 epochs during training.
- Data Testing: Evaluated model accuracy and performance using the testing dataset.
- Result Evaluation: Analyzed model accuracy and metrics.
- Data Division: 80% training, 20% testing
- Epochs: Trained for 60 epochs
- Accuracy: Reached 76% accuracy at epoch 55, dropped to 60-65% afterward
- Best Weight: Stored weights from epoch 55 with 76% precision, 77% recall, and 76% F1 score.
- Mean Average Precision (mAP@0.50): Achieved 76%
- Mean Average Precision (mAP@0.50): 76%
- Class-wise precision:
- Not-Used(westland): 80%
- Arable(cultivated): 73%
- None_arable (non-cultivated): 70%
The project involved collecting, preprocessing, annotating, and training a model using Google Earth images. The trained model achieved an overall mean average precision of 76%, with varying precision rates for different classes, notably 80% for wasteland, 73% for arable, and 70% for none_arable (non-cultivated). The highest accuracy was attained at epoch 55, and the model's performance slightly declined in subsequent epochs.
