remote_sensing module

RemoteSensingFeatures

A class for generating remote sensing features using TensorFlow.

This class provides static methods to compute various remote sensing indices and concatenate them into feature tensors.

Source code in aces/remote_sensing.py

class RemoteSensingFeatures:
    """
    A class for generating remote sensing features using TensorFlow.

    This class provides static methods to compute various remote sensing indices and concatenate them into feature tensors.
    """
    @staticmethod
    def normalized_difference(c1: tf.Tensor, c2: tf.Tensor, name: str = "nd") -> tf.Tensor:
        """
        Compute the normalized difference index between two spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the normalized difference index.

        """
        nd_f = keras.layers.Lambda(lambda x: ((x[0] - x[1]) / (x[0] + x[1])), name=name)([c1, c2])
        nd_inf = keras.layers.Lambda(lambda x: (x[0] - x[1]), name=f"{name}_inf")([c1, c2])
        return tf.where(tf.math.is_finite(nd_f), nd_f, nd_inf, name=name)

    @staticmethod
    def evi(c1: tf.Tensor, c2: tf.Tensor, c3: tf.Tensor, name: str = "evi") -> tf.Tensor:
        """
        Compute the enhanced vegetation index (EVI) using three spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            c3: A TensorFlow tensor representing the third spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the EVI.

        """
        _evi = keras.layers.Lambda(lambda x: 2.5 * ((x[0] - x[1]) / (x[0] + 6 * x[1] - 7.5 * x[2] + 1)), name=name)([c1, c2, c3])
        return _evi

    @staticmethod
    def savi(c1: tf.Tensor, c2: tf.Tensor, name: str = "savi") -> tf.Tensor:
        """
        Compute the soil-adjusted vegetation index (SAVI) between two spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the SAVI.

        """
        savi_f = keras.layers.Lambda(lambda x: ((x[0] - x[1]) / (x[0] + x[1] + 0.5)) * 1.5, name=name)([c1, c2])
        return savi_f

    @staticmethod
    def msavi(c1: tf.Tensor, c2: tf.Tensor, name: str = "msavi") -> tf.Tensor:
        """
        Compute the modified soil-adjusted vegetation index (MSAVI) between two spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the MSAVI.

        """
        msavi_f = keras.layers.Lambda(lambda x: (((2 * x[0] + 1) - tf.sqrt(((2 * x[0] + 1) * (2 * x[0] + 1)) - 8 * (x[0] - x[1]))) / 2), name=name)([c1, c2])
        return msavi_f

    @staticmethod
    def mtvi2(c1: tf.Tensor, c2: tf.Tensor, c3: tf.Tensor, name: str = "mtvi2") -> tf.Tensor:
        """
        Compute the modified transformed vegetation index 2 (MTVI2) using three spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            c3: A TensorFlow tensor representing the third spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the MTVI2.

        """
        mtvi2_f = keras.layers.Lambda(lambda x: (1.5 * (1.2 * (x[0] - x[2]) - 2.5 * (x[1] - x[2]))) / (tf.sqrt(((2 * x[0] + 1) * (2 * x[0] + 1)) - (6 * x[0] - 5 * tf.sqrt(x[1])) - 0.5)), name=name)([c1, c2, c3])
        return mtvi2_f

    @staticmethod
    def vari(c1: tf.Tensor, c2: tf.Tensor, c3: tf.Tensor, name: str = "vari") -> tf.Tensor:
        """
        Compute the visible atmospheric resistant index (VARI) using three spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            c3: A TensorFlow tensor representing the third spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the VARI.

        """
        vari_f = keras.layers.Lambda(lambda x: ((x[0] - x[1]) / (x[0] + x[1] - x[2])), name=name)([c1, c2, c3])
        vari_inf = keras.layers.Lambda(lambda x: (x[0] - x[1]), name=f"{name}_inf")([c1, c2, c3])
        return tf.where(tf.math.is_finite(vari_f), vari_f, vari_inf, name=name)

    @staticmethod
    def tgi(c1: tf.Tensor, c2: tf.Tensor, c3: tf.Tensor, name: str = "tgi") -> tf.Tensor:
        """
        Compute the triangular greenness index (TGI) using three spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            c3: A TensorFlow tensor representing the third spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the TGI.

        """
        tgi_f = keras.layers.Lambda(lambda x: ((120 * (x[1] - x[2])) - (190 * (x[1] - x[0]))) / 2, name=name)([c1, c2, c3])
        return tgi_f

    @staticmethod
    def ratio(c1: tf.Tensor, c2: tf.Tensor, name: str = "ratio") -> tf.Tensor:
        """
        Compute the ratio between two spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the ratio between the spectral bands.

        """
        ratio_f = keras.layers.Lambda(lambda x: x[0] / x[1], name=name)([c1, c2])
        ratio_inf = keras.layers.Lambda(lambda x: x[0], name=f"{name}_inf")([c1, c2])
        return tf.where(tf.math.is_finite(ratio_f), ratio_f, ratio_inf, name=name)

    @staticmethod
    def nvi(c1: tf.Tensor, c2: tf.Tensor, name: str = "nvi") -> tf.Tensor:
        """
        Compute the normalized vegetation index (NVI) between two spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the NVI.

        """
        nvi_f = keras.layers.Lambda(lambda x: x[0] / (x[0] + x[1]), name=name)([c1, c2])
        nvi_inf = keras.layers.Lambda(lambda x: x[0], name=f"{name}_inf")([c1, c2])
        return tf.where(tf.math.is_finite(nvi_f), nvi_f, nvi_inf, name=name)

    @staticmethod
    def diff_band(c1: tf.Tensor, c2: tf.Tensor, name: str = "diff") -> tf.Tensor:
        """
        Compute the difference between two spectral bands.

        Args:
            c1: A TensorFlow tensor representing the first spectral band.
            c2: A TensorFlow tensor representing the second spectral band.
            name: A string specifying the name for the operation.

        Returns:
            A TensorFlow tensor representing the difference between the spectral bands.

        """
        diff = keras.layers.Lambda(lambda x: x[0] - x[1], name=name)([c1, c2])
        return diff

    @staticmethod
    def concatenate_features_for_cnn(input_tensor: tf.Tensor) -> tf.Tensor:
        """
        Concatenate remote sensing features for Convolutional Neural Network (CNN) input.

        Args:
            input_tensor: A TensorFlow tensor representing the input remote sensing data.

        Returns:
            A TensorFlow tensor representing the concatenated features for CNN input.

        """
        red_before = input_tensor[:, :, :, 0:1]
        green_before = input_tensor[:, :, :, 1:2]
        blue_before = input_tensor[:, :, :, 2:3]
        nir_before = input_tensor[:, :, :, 3:4]
        red_during = input_tensor[:, :, :, 4:5]
        green_during = input_tensor[:, :, :, 5:6]
        blue_during = input_tensor[:, :, :, 6:7]
        nir_during = input_tensor[:, :, :, 7:8]

        ndvi_before = RemoteSensingFeatures.normalized_difference(nir_before, red_before, name="ndvi_before")
        ndvi_during = RemoteSensingFeatures.normalized_difference(nir_during, red_during, name="ndvi_during")
        evi_before = RemoteSensingFeatures.evi(nir_before, red_before, blue_before, name="evi_before")
        evi_during = RemoteSensingFeatures.evi(nir_during, red_during, blue_during, name="evi_during")
        ndwi_before = RemoteSensingFeatures.normalized_difference(green_before, nir_before, name="ndwi_before")
        ndwi_during = RemoteSensingFeatures.normalized_difference(green_during, nir_during, name="ndwi_during")
        savi_before = RemoteSensingFeatures.savi(nir_before, red_before, name="savi_before")
        savi_during = RemoteSensingFeatures.savi(nir_during, red_during, name="savi_during")
        # msavi_before = RemoteSensingFeatures.msavi(nir_before, red_before, name="msavi_before")
        # msavi_during = RemoteSensingFeatures.msavi(nir_during, red_during, name="msavi_during")
        mtvi2_before = RemoteSensingFeatures.mtvi2(nir_before, red_before, green_before, name="mtvi2_before")
        mtvi2_during = RemoteSensingFeatures.mtvi2(nir_during, red_during, green_during, name="mtvi2_during")
        # vari is not used because the computation gave some nan values
        # vari_before = RemoteSensingFeatures.vari(green_before, red_before, blue_before, name="vari_before")
        # vari_during = RemoteSensingFeatures.vari(green_during, red_during, blue_during, name="vari_during")
        # tgi_before = RemoteSensingFeatures.tgi(green_before, red_before, blue_before, name="tgi_before")
        # tgi_during = RemoteSensingFeatures.tgi(green_during, red_during, blue_during, name="tgi_during")

        red_diff1 = RemoteSensingFeatures.diff_band(red_before, red_during, name="diff1")
        green_diff1 = RemoteSensingFeatures.diff_band(green_before, green_during, name="diff2")
        blue_diff1 = RemoteSensingFeatures.diff_band(blue_before, blue_during, name="diff3")
        nir_diff1 = RemoteSensingFeatures.diff_band(nir_before, nir_during, name="diff4")

        return keras.layers.concatenate(
            [input_tensor, ndvi_before, ndvi_during, evi_before, evi_during, ndwi_before, ndwi_during, savi_before, savi_during,
             mtvi2_before, mtvi2_during,
             red_diff1, green_diff1, blue_diff1, nir_diff1],
            name="input_features"
        )

    @staticmethod
    def concatenate_features_for_dnn(input_tensor: tf.Tensor) -> tf.Tensor:
        """
        Concatenate remote sensing features for Deep Neural Network (DNN) input.

        Args:
            input_tensor: A TensorFlow tensor representing the input remote sensing data.

        Returns:
            A TensorFlow tensor representing the concatenated features for DNN input.

        """
        red_before = input_tensor[:, :, :, 0:1]
        green_before = input_tensor[:, :, :, 1:2]
        blue_before = input_tensor[:, :, :, 2:3]
        nir_before = input_tensor[:, :, :, 3:4]
        red_during = input_tensor[:, :, :, 4:5]
        green_during = input_tensor[:, :, :, 5:6]
        blue_during = input_tensor[:, :, :, 6:7]
        nir_during = input_tensor[:, :, :, 7:8]

        ndvi_before = RemoteSensingFeatures.normalized_difference(nir_before, red_before, name="ndvi_before")
        ndvi_during = RemoteSensingFeatures.normalized_difference(nir_during, red_during, name="ndvi_during")
        evi_before = RemoteSensingFeatures.evi(nir_before, red_before, blue_before, name="evi_before")
        evi_during = RemoteSensingFeatures.evi(nir_during, red_during, blue_during, name="evi_during")
        ndwi_before = RemoteSensingFeatures.normalized_difference(green_before, nir_before, name="ndwi_before")
        ndwi_during = RemoteSensingFeatures.normalized_difference(green_during, nir_during, name="ndwi_during")
        savi_before = RemoteSensingFeatures.savi(nir_before, red_before, name="savi_before")
        savi_during = RemoteSensingFeatures.savi(nir_during, red_during, name="savi_during")
        msavi_before = RemoteSensingFeatures.msavi(nir_before, red_before, name="msavi_before")
        msavi_during = RemoteSensingFeatures.msavi(nir_during, red_during, name="msavi_during")
        mtvi2_before = RemoteSensingFeatures.mtvi2(nir_before, red_before, green_before, name="mtvi2_before")
        mtvi2_during = RemoteSensingFeatures.mtvi2(nir_during, red_during, green_during, name="mtvi2_during")
        vari_before = RemoteSensingFeatures.vari(green_before, red_before, blue_before, name="vari_before")
        vari_during = RemoteSensingFeatures.vari(green_during, red_during, blue_during, name="vari_during")
        tgi_before = RemoteSensingFeatures.tgi(green_before, red_before, blue_before, name="tgi_before")
        tgi_during = RemoteSensingFeatures.tgi(green_during, red_during, blue_during, name="tgi_during")

        return keras.layers.concatenate(
            [input_tensor, ndvi_before, ndvi_during, evi_before, evi_during, ndwi_before, ndwi_during, savi_before, savi_during,
             msavi_before, msavi_during, mtvi2_before, mtvi2_during, vari_before, vari_during, tgi_before, tgi_during],
            name="input_features"
        )

    @staticmethod
    def derive_features_for_dnn(features_dict: dict, added_features: list = []) -> dict:
        """
        Concatenate remote sensing features for Deep Neural Network (DNN) input.

        Args:
            input_tensor: A TensorFlow tensor representing the input remote sensing data.

        Returns:
            A TensorFlow tensor representing the concatenated features for DNN input.

        """
        red_before = features_dict["red_before"]
        green_before = features_dict["green_before"]
        blue_before = features_dict["blue_before"]
        nir_before = features_dict["nir_before"]
        red_during = features_dict["red_during"]
        green_during = features_dict["green_during"]
        blue_during = features_dict["blue_during"]
        nir_during = features_dict["nir_during"]

        feature_col = {
            "ndvi_before": RemoteSensingFeatures.normalized_difference(nir_before, red_before, name="ndvi_before"),
            "ndvi_during": RemoteSensingFeatures.normalized_difference(nir_during, red_during, name="ndvi_during"),
            "evi_before": RemoteSensingFeatures.evi(nir_before, red_before, blue_before, name="evi_before"),
            "evi_during": RemoteSensingFeatures.evi(nir_during, red_during, blue_during, name="evi_during"),
            "ndwi_before": RemoteSensingFeatures.normalized_difference(green_before, nir_before, name="ndwi_before"),
            "ndwi_during": RemoteSensingFeatures.normalized_difference(green_during, nir_during, name="ndwi_during"),
            "savi_before": RemoteSensingFeatures.savi(nir_before, red_before, name="savi_before"),
            "savi_during": RemoteSensingFeatures.savi(nir_during, red_during, name="savi_during"),
            "msavi_before": RemoteSensingFeatures.msavi(nir_before, red_before, name="msavi_before"),
            "msavi_during": RemoteSensingFeatures.msavi(nir_during, red_during, name="msavi_during"),
            "mtvi2_before": RemoteSensingFeatures.mtvi2(nir_before, red_before, green_before, name="mtvi2_before"),
            "mtvi2_during": RemoteSensingFeatures.mtvi2(nir_during, red_during, green_during, name="mtvi2_during"),
            "vari_before": RemoteSensingFeatures.vari(green_before, red_before, blue_before, name="vari_before"),
            "vari_during": RemoteSensingFeatures.vari(green_during, red_during, blue_during, name="vari_during"),
            "tgi_before": RemoteSensingFeatures.tgi(green_before, red_before, blue_before, name="tgi_before"),
            "tgi_during": RemoteSensingFeatures.tgi(green_during, red_during, blue_during, name="tgi_during"),
        }

        for feature in added_features:
            features_dict[feature] = feature_col[feature]

        return features_dict

concatenate_features_for_cnn(input_tensor) staticmethod

Concatenate remote sensing features for Convolutional Neural Network (CNN) input.

Parameters:

Name	Type	Description	Default
input_tensor	Tensor	A TensorFlow tensor representing the input remote sensing data.	required

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the concatenated features for CNN input.

Source code in aces/remote_sensing.py

@staticmethod
def concatenate_features_for_cnn(input_tensor: tf.Tensor) -> tf.Tensor:
    """
    Concatenate remote sensing features for Convolutional Neural Network (CNN) input.

    Args:
        input_tensor: A TensorFlow tensor representing the input remote sensing data.

    Returns:
        A TensorFlow tensor representing the concatenated features for CNN input.

    """
    red_before = input_tensor[:, :, :, 0:1]
    green_before = input_tensor[:, :, :, 1:2]
    blue_before = input_tensor[:, :, :, 2:3]
    nir_before = input_tensor[:, :, :, 3:4]
    red_during = input_tensor[:, :, :, 4:5]
    green_during = input_tensor[:, :, :, 5:6]
    blue_during = input_tensor[:, :, :, 6:7]
    nir_during = input_tensor[:, :, :, 7:8]

    ndvi_before = RemoteSensingFeatures.normalized_difference(nir_before, red_before, name="ndvi_before")
    ndvi_during = RemoteSensingFeatures.normalized_difference(nir_during, red_during, name="ndvi_during")
    evi_before = RemoteSensingFeatures.evi(nir_before, red_before, blue_before, name="evi_before")
    evi_during = RemoteSensingFeatures.evi(nir_during, red_during, blue_during, name="evi_during")
    ndwi_before = RemoteSensingFeatures.normalized_difference(green_before, nir_before, name="ndwi_before")
    ndwi_during = RemoteSensingFeatures.normalized_difference(green_during, nir_during, name="ndwi_during")
    savi_before = RemoteSensingFeatures.savi(nir_before, red_before, name="savi_before")
    savi_during = RemoteSensingFeatures.savi(nir_during, red_during, name="savi_during")
    # msavi_before = RemoteSensingFeatures.msavi(nir_before, red_before, name="msavi_before")
    # msavi_during = RemoteSensingFeatures.msavi(nir_during, red_during, name="msavi_during")
    mtvi2_before = RemoteSensingFeatures.mtvi2(nir_before, red_before, green_before, name="mtvi2_before")
    mtvi2_during = RemoteSensingFeatures.mtvi2(nir_during, red_during, green_during, name="mtvi2_during")
    # vari is not used because the computation gave some nan values
    # vari_before = RemoteSensingFeatures.vari(green_before, red_before, blue_before, name="vari_before")
    # vari_during = RemoteSensingFeatures.vari(green_during, red_during, blue_during, name="vari_during")
    # tgi_before = RemoteSensingFeatures.tgi(green_before, red_before, blue_before, name="tgi_before")
    # tgi_during = RemoteSensingFeatures.tgi(green_during, red_during, blue_during, name="tgi_during")

    red_diff1 = RemoteSensingFeatures.diff_band(red_before, red_during, name="diff1")
    green_diff1 = RemoteSensingFeatures.diff_band(green_before, green_during, name="diff2")
    blue_diff1 = RemoteSensingFeatures.diff_band(blue_before, blue_during, name="diff3")
    nir_diff1 = RemoteSensingFeatures.diff_band(nir_before, nir_during, name="diff4")

    return keras.layers.concatenate(
        [input_tensor, ndvi_before, ndvi_during, evi_before, evi_during, ndwi_before, ndwi_during, savi_before, savi_during,
         mtvi2_before, mtvi2_during,
         red_diff1, green_diff1, blue_diff1, nir_diff1],
        name="input_features"
    )

concatenate_features_for_dnn(input_tensor) staticmethod

Concatenate remote sensing features for Deep Neural Network (DNN) input.

Parameters:

Name	Type	Description	Default
input_tensor	Tensor	A TensorFlow tensor representing the input remote sensing data.	required

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the concatenated features for DNN input.

Source code in aces/remote_sensing.py

@staticmethod
def concatenate_features_for_dnn(input_tensor: tf.Tensor) -> tf.Tensor:
    """
    Concatenate remote sensing features for Deep Neural Network (DNN) input.

    Args:
        input_tensor: A TensorFlow tensor representing the input remote sensing data.

    Returns:
        A TensorFlow tensor representing the concatenated features for DNN input.

    """
    red_before = input_tensor[:, :, :, 0:1]
    green_before = input_tensor[:, :, :, 1:2]
    blue_before = input_tensor[:, :, :, 2:3]
    nir_before = input_tensor[:, :, :, 3:4]
    red_during = input_tensor[:, :, :, 4:5]
    green_during = input_tensor[:, :, :, 5:6]
    blue_during = input_tensor[:, :, :, 6:7]
    nir_during = input_tensor[:, :, :, 7:8]

    ndvi_before = RemoteSensingFeatures.normalized_difference(nir_before, red_before, name="ndvi_before")
    ndvi_during = RemoteSensingFeatures.normalized_difference(nir_during, red_during, name="ndvi_during")
    evi_before = RemoteSensingFeatures.evi(nir_before, red_before, blue_before, name="evi_before")
    evi_during = RemoteSensingFeatures.evi(nir_during, red_during, blue_during, name="evi_during")
    ndwi_before = RemoteSensingFeatures.normalized_difference(green_before, nir_before, name="ndwi_before")
    ndwi_during = RemoteSensingFeatures.normalized_difference(green_during, nir_during, name="ndwi_during")
    savi_before = RemoteSensingFeatures.savi(nir_before, red_before, name="savi_before")
    savi_during = RemoteSensingFeatures.savi(nir_during, red_during, name="savi_during")
    msavi_before = RemoteSensingFeatures.msavi(nir_before, red_before, name="msavi_before")
    msavi_during = RemoteSensingFeatures.msavi(nir_during, red_during, name="msavi_during")
    mtvi2_before = RemoteSensingFeatures.mtvi2(nir_before, red_before, green_before, name="mtvi2_before")
    mtvi2_during = RemoteSensingFeatures.mtvi2(nir_during, red_during, green_during, name="mtvi2_during")
    vari_before = RemoteSensingFeatures.vari(green_before, red_before, blue_before, name="vari_before")
    vari_during = RemoteSensingFeatures.vari(green_during, red_during, blue_during, name="vari_during")
    tgi_before = RemoteSensingFeatures.tgi(green_before, red_before, blue_before, name="tgi_before")
    tgi_during = RemoteSensingFeatures.tgi(green_during, red_during, blue_during, name="tgi_during")

    return keras.layers.concatenate(
        [input_tensor, ndvi_before, ndvi_during, evi_before, evi_during, ndwi_before, ndwi_during, savi_before, savi_during,
         msavi_before, msavi_during, mtvi2_before, mtvi2_during, vari_before, vari_during, tgi_before, tgi_during],
        name="input_features"
    )

derive_features_for_dnn(features_dict, added_features=[]) staticmethod

Concatenate remote sensing features for Deep Neural Network (DNN) input.

Parameters:

Name	Type	Description	Default
input_tensor		A TensorFlow tensor representing the input remote sensing data.	required

Returns:

Type	Description
dict	A TensorFlow tensor representing the concatenated features for DNN input.

Source code in aces/remote_sensing.py

@staticmethod
def derive_features_for_dnn(features_dict: dict, added_features: list = []) -> dict:
    """
    Concatenate remote sensing features for Deep Neural Network (DNN) input.

    Args:
        input_tensor: A TensorFlow tensor representing the input remote sensing data.

    Returns:
        A TensorFlow tensor representing the concatenated features for DNN input.

    """
    red_before = features_dict["red_before"]
    green_before = features_dict["green_before"]
    blue_before = features_dict["blue_before"]
    nir_before = features_dict["nir_before"]
    red_during = features_dict["red_during"]
    green_during = features_dict["green_during"]
    blue_during = features_dict["blue_during"]
    nir_during = features_dict["nir_during"]

    feature_col = {
        "ndvi_before": RemoteSensingFeatures.normalized_difference(nir_before, red_before, name="ndvi_before"),
        "ndvi_during": RemoteSensingFeatures.normalized_difference(nir_during, red_during, name="ndvi_during"),
        "evi_before": RemoteSensingFeatures.evi(nir_before, red_before, blue_before, name="evi_before"),
        "evi_during": RemoteSensingFeatures.evi(nir_during, red_during, blue_during, name="evi_during"),
        "ndwi_before": RemoteSensingFeatures.normalized_difference(green_before, nir_before, name="ndwi_before"),
        "ndwi_during": RemoteSensingFeatures.normalized_difference(green_during, nir_during, name="ndwi_during"),
        "savi_before": RemoteSensingFeatures.savi(nir_before, red_before, name="savi_before"),
        "savi_during": RemoteSensingFeatures.savi(nir_during, red_during, name="savi_during"),
        "msavi_before": RemoteSensingFeatures.msavi(nir_before, red_before, name="msavi_before"),
        "msavi_during": RemoteSensingFeatures.msavi(nir_during, red_during, name="msavi_during"),
        "mtvi2_before": RemoteSensingFeatures.mtvi2(nir_before, red_before, green_before, name="mtvi2_before"),
        "mtvi2_during": RemoteSensingFeatures.mtvi2(nir_during, red_during, green_during, name="mtvi2_during"),
        "vari_before": RemoteSensingFeatures.vari(green_before, red_before, blue_before, name="vari_before"),
        "vari_during": RemoteSensingFeatures.vari(green_during, red_during, blue_during, name="vari_during"),
        "tgi_before": RemoteSensingFeatures.tgi(green_before, red_before, blue_before, name="tgi_before"),
        "tgi_during": RemoteSensingFeatures.tgi(green_during, red_during, blue_during, name="tgi_during"),
    }

    for feature in added_features:
        features_dict[feature] = feature_col[feature]

    return features_dict

diff_band(c1, c2, name='diff') staticmethod

Compute the difference between two spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
name	str	A string specifying the name for the operation.	'diff'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the difference between the spectral bands.

Source code in aces/remote_sensing.py

@staticmethod
def diff_band(c1: tf.Tensor, c2: tf.Tensor, name: str = "diff") -> tf.Tensor:
    """
    Compute the difference between two spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the difference between the spectral bands.

    """
    diff = keras.layers.Lambda(lambda x: x[0] - x[1], name=name)([c1, c2])
    return diff

evi(c1, c2, c3, name='evi') staticmethod

Compute the enhanced vegetation index (EVI) using three spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
c3	Tensor	A TensorFlow tensor representing the third spectral band.	required
name	str	A string specifying the name for the operation.	'evi'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the EVI.

Source code in aces/remote_sensing.py

@staticmethod
def evi(c1: tf.Tensor, c2: tf.Tensor, c3: tf.Tensor, name: str = "evi") -> tf.Tensor:
    """
    Compute the enhanced vegetation index (EVI) using three spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        c3: A TensorFlow tensor representing the third spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the EVI.

    """
    _evi = keras.layers.Lambda(lambda x: 2.5 * ((x[0] - x[1]) / (x[0] + 6 * x[1] - 7.5 * x[2] + 1)), name=name)([c1, c2, c3])
    return _evi

msavi(c1, c2, name='msavi') staticmethod

Compute the modified soil-adjusted vegetation index (MSAVI) between two spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
name	str	A string specifying the name for the operation.	'msavi'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the MSAVI.

Source code in aces/remote_sensing.py

@staticmethod
def msavi(c1: tf.Tensor, c2: tf.Tensor, name: str = "msavi") -> tf.Tensor:
    """
    Compute the modified soil-adjusted vegetation index (MSAVI) between two spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the MSAVI.

    """
    msavi_f = keras.layers.Lambda(lambda x: (((2 * x[0] + 1) - tf.sqrt(((2 * x[0] + 1) * (2 * x[0] + 1)) - 8 * (x[0] - x[1]))) / 2), name=name)([c1, c2])
    return msavi_f

mtvi2(c1, c2, c3, name='mtvi2') staticmethod

Compute the modified transformed vegetation index 2 (MTVI2) using three spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
c3	Tensor	A TensorFlow tensor representing the third spectral band.	required
name	str	A string specifying the name for the operation.	'mtvi2'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the MTVI2.

Source code in aces/remote_sensing.py

@staticmethod
def mtvi2(c1: tf.Tensor, c2: tf.Tensor, c3: tf.Tensor, name: str = "mtvi2") -> tf.Tensor:
    """
    Compute the modified transformed vegetation index 2 (MTVI2) using three spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        c3: A TensorFlow tensor representing the third spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the MTVI2.

    """
    mtvi2_f = keras.layers.Lambda(lambda x: (1.5 * (1.2 * (x[0] - x[2]) - 2.5 * (x[1] - x[2]))) / (tf.sqrt(((2 * x[0] + 1) * (2 * x[0] + 1)) - (6 * x[0] - 5 * tf.sqrt(x[1])) - 0.5)), name=name)([c1, c2, c3])
    return mtvi2_f

normalized_difference(c1, c2, name='nd') staticmethod

Compute the normalized difference index between two spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
name	str	A string specifying the name for the operation.	'nd'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the normalized difference index.

Source code in aces/remote_sensing.py

@staticmethod
def normalized_difference(c1: tf.Tensor, c2: tf.Tensor, name: str = "nd") -> tf.Tensor:
    """
    Compute the normalized difference index between two spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the normalized difference index.

    """
    nd_f = keras.layers.Lambda(lambda x: ((x[0] - x[1]) / (x[0] + x[1])), name=name)([c1, c2])
    nd_inf = keras.layers.Lambda(lambda x: (x[0] - x[1]), name=f"{name}_inf")([c1, c2])
    return tf.where(tf.math.is_finite(nd_f), nd_f, nd_inf, name=name)

nvi(c1, c2, name='nvi') staticmethod

Compute the normalized vegetation index (NVI) between two spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
name	str	A string specifying the name for the operation.	'nvi'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the NVI.

Source code in aces/remote_sensing.py

@staticmethod
def nvi(c1: tf.Tensor, c2: tf.Tensor, name: str = "nvi") -> tf.Tensor:
    """
    Compute the normalized vegetation index (NVI) between two spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the NVI.

    """
    nvi_f = keras.layers.Lambda(lambda x: x[0] / (x[0] + x[1]), name=name)([c1, c2])
    nvi_inf = keras.layers.Lambda(lambda x: x[0], name=f"{name}_inf")([c1, c2])
    return tf.where(tf.math.is_finite(nvi_f), nvi_f, nvi_inf, name=name)

ratio(c1, c2, name='ratio') staticmethod

Compute the ratio between two spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
name	str	A string specifying the name for the operation.	'ratio'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the ratio between the spectral bands.

Source code in aces/remote_sensing.py

@staticmethod
def ratio(c1: tf.Tensor, c2: tf.Tensor, name: str = "ratio") -> tf.Tensor:
    """
    Compute the ratio between two spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the ratio between the spectral bands.

    """
    ratio_f = keras.layers.Lambda(lambda x: x[0] / x[1], name=name)([c1, c2])
    ratio_inf = keras.layers.Lambda(lambda x: x[0], name=f"{name}_inf")([c1, c2])
    return tf.where(tf.math.is_finite(ratio_f), ratio_f, ratio_inf, name=name)

savi(c1, c2, name='savi') staticmethod

Compute the soil-adjusted vegetation index (SAVI) between two spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
name	str	A string specifying the name for the operation.	'savi'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the SAVI.

Source code in aces/remote_sensing.py

@staticmethod
def savi(c1: tf.Tensor, c2: tf.Tensor, name: str = "savi") -> tf.Tensor:
    """
    Compute the soil-adjusted vegetation index (SAVI) between two spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the SAVI.

    """
    savi_f = keras.layers.Lambda(lambda x: ((x[0] - x[1]) / (x[0] + x[1] + 0.5)) * 1.5, name=name)([c1, c2])
    return savi_f

tgi(c1, c2, c3, name='tgi') staticmethod

Compute the triangular greenness index (TGI) using three spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
c3	Tensor	A TensorFlow tensor representing the third spectral band.	required
name	str	A string specifying the name for the operation.	'tgi'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the TGI.

Source code in aces/remote_sensing.py

@staticmethod
def tgi(c1: tf.Tensor, c2: tf.Tensor, c3: tf.Tensor, name: str = "tgi") -> tf.Tensor:
    """
    Compute the triangular greenness index (TGI) using three spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        c3: A TensorFlow tensor representing the third spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the TGI.

    """
    tgi_f = keras.layers.Lambda(lambda x: ((120 * (x[1] - x[2])) - (190 * (x[1] - x[0]))) / 2, name=name)([c1, c2, c3])
    return tgi_f

vari(c1, c2, c3, name='vari') staticmethod

Compute the visible atmospheric resistant index (VARI) using three spectral bands.

Parameters:

Name	Type	Description	Default
c1	Tensor	A TensorFlow tensor representing the first spectral band.	required
c2	Tensor	A TensorFlow tensor representing the second spectral band.	required
c3	Tensor	A TensorFlow tensor representing the third spectral band.	required
name	str	A string specifying the name for the operation.	'vari'

Returns:

Type	Description
Tensor	A TensorFlow tensor representing the VARI.

Source code in aces/remote_sensing.py

@staticmethod
def vari(c1: tf.Tensor, c2: tf.Tensor, c3: tf.Tensor, name: str = "vari") -> tf.Tensor:
    """
    Compute the visible atmospheric resistant index (VARI) using three spectral bands.

    Args:
        c1: A TensorFlow tensor representing the first spectral band.
        c2: A TensorFlow tensor representing the second spectral band.
        c3: A TensorFlow tensor representing the third spectral band.
        name: A string specifying the name for the operation.

    Returns:
        A TensorFlow tensor representing the VARI.

    """
    vari_f = keras.layers.Lambda(lambda x: ((x[0] - x[1]) / (x[0] + x[1] - x[2])), name=name)([c1, c2, c3])
    vari_inf = keras.layers.Lambda(lambda x: (x[0] - x[1]), name=f"{name}_inf")([c1, c2, c3])
    return tf.where(tf.math.is_finite(vari_f), vari_f, vari_inf, name=name)