diff --git a/README.md b/README.md
index a815482..52a037b 100644
--- a/README.md
+++ b/README.md
@@ -30,4 +30,27 @@ It gets the data matrix from a supported CSV format from either a VFS or downloa
 ```
 from cover_class.static.retrieval import generate_hdf5_from_config
 generate_hdf5_from_config('/path/to/my/config.yml')
-```
\ No newline at end of file
+```
+
+### Outlier Detection
+A separate feature of the cover-class repository is the ability to utilize outlier detectors which will save out a png highlighting any outliers and provide the indices in the dataset of them. 
+
+The current options available are:
+- z-score
+- kmeans
+- mahalanobis
+- lof (Local Outlier Factor)
+
+Example:
+```
+>>> import numpy as np
+>>> my_data = np.load('my_data.npy')
+>>> 
+>>> from cover_class.outlier_detection import show_outliers
+>>> kwargs = {'outlier_percentile': 80}
+>>> show_outliers(my_data, 'mahalanobis', png_name='my-data-outliers.png', **kwargs)
+array([ 2,  4,  5, 37, 50, 59, 60])
+```
+
+And an example of an output png:
+![outlier detection figure](figs/outlier-detection-readme-fig.png)
\ No newline at end of file
diff --git a/figs/outlier-detection-readme-fig.png b/figs/outlier-detection-readme-fig.png
new file mode 100644
index 0000000..2ced964
Binary files /dev/null and b/figs/outlier-detection-readme-fig.png differ
diff --git a/src/cover_class/outlier_detection.py b/src/cover_class/outlier_detection.py
new file mode 100644
index 0000000..0ea81d1
--- /dev/null
+++ b/src/cover_class/outlier_detection.py
@@ -0,0 +1,52 @@
+from numpy.typing import NDArray
+import numpy as np
+from scipy.stats import zscore #type: ignore
+from sklearn.neighbors import NearestNeighbors, LocalOutlierFactor #type: ignore
+from scipy.spatial.distance import mahalanobis #type: ignore
+import matplotlib.pyplot as plt
+
+def zcore_outliers(data:NDArray, **kwargs) -> NDArray:
+    z = np.abs(zscore(data))
+    return (z > 3).any(axis=1)
+
+def kmeans_outliers(data:NDArray, outlier_percentile:int=95, **kwargs) -> NDArray:
+    nbrs = NearestNeighbors(n_neighbors=5).fit(data)
+    dists, _ = nbrs.kneighbors(data)
+    return dists.mean(axis=1) > np.percentile(dists.mean(axis=1), outlier_percentile)
+
+def mahalanobis_distance(data:NDArray, outlier_percentile:int=95, **kwargs) -> NDArray:
+    cov = np.cov(data.T)
+    inv_cov = np.linalg.pinv(cov)
+    center = data.mean(0)
+    m = np.array([mahalanobis(x, center, inv_cov) for x in data])
+    return m > np.percentile(m, outlier_percentile)
+
+def local_outlier_factor(data:NDArray, metric='cosine', **kwargs) -> NDArray:
+    return LocalOutlierFactor(metric=metric, **kwargs).fit_predict(data) == -1
+
+def show_outliers(data:NDArray, method:str='z-score', png_name:str='', **kwargs) -> NDArray:
+    # returns the indices of outliers if there are any
+    outliers = np.ndarray([], dtype=bool)
+    match method:
+        case'z-score':
+            outliers = zcore_outliers(data, **kwargs)
+        case 'kmeans':
+            outliers = kmeans_outliers(data, **kwargs)
+        case 'mahalanobis':
+            outliers = mahalanobis_distance(data, **kwargs)
+        case 'lof':
+            outliers = local_outlier_factor(data, **kwargs)
+        case _:
+            raise ValueError('Unsupported outlier method: '+method)
+    if outliers.sum() == 0:
+        return np.array([])
+    
+    plt.figure(figsize=(8, 5))
+    plt.plot(data[~outliers].T, color='black', alpha=0.1)
+    plt.plot(data[outliers].T, color='red', alpha=0.9)
+    plt.title('Outliers using '+method)
+    if png_name != '':
+        plt.savefig(png_name)
+    plt.show()
+
+    return np.where(outliers)[0] 
diff --git a/src/cover_class/train.py b/src/cover_class/train.py
index a9d6786..de35647 100644
--- a/src/cover_class/train.py
+++ b/src/cover_class/train.py
@@ -51,7 +51,7 @@ def setup_training_from_config(
     odl = dataloader_from_config(
         config, 
         FloatTensor(train_spectra),
-        train_labels,
+        LongTensor(train_labels.to(dtype=torch.long)),
         batch_size,
         shuffle,
     )