Examples

Note

• You will have to download the datasets from their respective source. You can find a link for each dataset in the respective dataset class in aitlas/datasets/ or use the AiTLAS Semantic Data Catalog available at eodata.bvlabs.ai .

• You can also find various trained models, model configurations and processing details of many datasets (with their corresponding splits used for training and evaluating the models) in our AiTLAS Benchmark Arena repository available at aitlas-arena.bvlabs.ai .

Multi-class classification tasks

• Multiclass Dataset statistics
  • AID dataset
  • Eurosat dataset
  • PatternNet dataset
  • Resisc45 dataset
  • RSD46WHU dataset train split
  • RSD46WHU dataset test split
  • RSICB256 dataset
  • RSSCN7 dataset
  • SAT6 train split
  • SAT6 test split
  • SiriWhu dataset
  • UcMerced dataset
  • WHURS19 dataset
  • CLRS dataset
  • Brazilian Coffee Scenes dataset
  • Optimal 31 Dataset
  • So2Sat Dataset train split
  • So2Sat Dataset validation split
  • So2Sat Dataset test split
• AID Dataset: Example of the aitlas toolbox for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• CLRS Dataset: Example of the aitlas toolbox for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and validation splits
  • Setup and create the model for training
  • Training and validation
  • Test the model
  • Predictions
• EuroSAT dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• PatterNet dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• Resisc45 dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• RSD46 dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• RSI CB256 dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• RSSCN7 dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• SAT6 dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• Siri Whu dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• UCI Merced dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• WHU-RS19 dataset: Example of the aitlas toolbox in for multi class image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions

Multi-label classification tasks

• Multi Label datasests statistics
  • AID dataset train split
  • AID dataset test split
  • DFC15 dataset train split
  • DFC15 dataset test split
  • UC Merced dataset
  • MLRSNet dataset
  • Planet UAS dataset
  • Big Earth Net Dataset with 19 labels
  • Big Earth Net Dataset with 43 labels
• AID mlc dataset: Example of the aitlas toolbox for multi label image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• BigEarth Net 19 dataset: Example of the aitlas toolbox for multi label image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• BigEarth Net 43 dataset: Example of the aitlas toolbox for multi label image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• DFC15 dataset: Example of the aitlas toolbox for multi label image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• MLRS Net dataset: Example of the aitlas toolbox for multi label image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• Planet UAS dataset: Example of the aitlas toolbox for multi label image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions
• UC Merced dataset: Example of the aitlas toolbox for multi label image classification
  • Load the dataset
  • Show images from the dataset
  • Inspect the data
  • Load train and test splits
  • Setup and create the model for training
  • Training and evaluation
  • Predictions

Visualization of model predictions with GradCam

• GradCam example
  • Load and transform the input image
• GradCAM for classification using ResNet50 model
• GradCAM for classification using ViT model

Semantic image segmentation tasks

• Semantic image segmentation dataset statistics
  • Massachusetts Buildings Dataset
  • Massachusetts Roads Dataset
  • Landcover AI Dataset
  • Chactun Dataset
  • Inria Dataset
  • Amazon Rain Forest Dataset
  • AIRS (Aerial Imagery for Roof Segmentation) Dataset
• AIRS Dataset: Example of the aitlas toolbox for semantic segmentation
  • Import the required packages
  • Visualize images and masks
  • Load train data
  • Create the model
  • Start the training
  • Evalute the model using test data
  • Predictions
• Amazon Forest dataset: Example of the aitlas toolbox for semantic segmentation
  • Import the required packages
  • Visualize images and masks
  • Load train data
  • Create the model
  • Start the training
  • Evalute the model using test data
  • Predictions
• CamVid dataset: Example of the aitlas toolbox for semantic segmentation
  • Import the required packages
  • Visualize images and masks
  • Load train data
  • Create the model
  • Start the training
  • Evalute the model using test data
  • Predictions
• Chactun dataset: Example of the aitlas toolbox for image segmentation
  • Import the required packages
  • Visualize images and masks
  • Load train data
  • Create the model
  • Start the training
  • Evalute the model using test data
  • Predictions
• Inria dataset: dataset: Example of the aitlas toolbox for image segmentation
  • Import the required packages
  • Visualize images and masks
  • Load train data
  • Create the model
  • Start the training
  • Predictions
• LandCover AI dataset: Example of the aitlas toolbox in the context of image segmentation
  • Import the required packages
  • Visualize images and masks
  • Load train data
  • Create the model
  • Start the training
  • Evalute the model using test data
  • Predictions
• MassachusettsBuildings dataset: Example of the aitlas toolbox in the context of image segmentation
  • Import the required packages
  • Visualize images and masks
  • Load train data
  • Create the model
  • Start the training
  • Evalute the model using test data
  • Predictions
• Massachusetts Roads dataset: Example of the aitlas toolbox in the context of image segmentation
  • Import the required packages
  • Visualize images and masks
  • Load train data
  • Create the model
  • Start the training
  • Evalute the model using test data
  • Predictions

Object detection tasks

• Example of the aitlas toolbox in the context of object detection in remote sensing images
  • Import the required packages
  • Visualize images and objects witn bounding boxes
  • Load train data
  • Create the model
  • Start the training
  • Evalute the model using test data
  • Predictions

Land-use classification tasks

• Example of the aitlas toolbox for benchmarking a dataset
  • Define the splits and dataset paths
  • Loop through the splits, train and evaluate
  • See the results

Crop-type prediction tasks

• Example of the aitlas toolbox for crop type prediction
  • Load the dataset
  • Visualize sample time series from the dataset
  • Inspect the data
  • Load train and validation splits
  • Setup and create the model for training
  • Training and validation
  • Test the model
  • Running and visualizing predictions

EO patch data analysis (Breizhcrops example)

• Read Breizhcrops EOPatch files
• Model Evaluation on Breizhcrops dataset
  • Dataset split:
    • Explore train dataset
    • Explore test dataset
• Training the models
  • Compile L1C Results