How to load and save 3D Numpy Array file using savetxt() and loadtxt() functions?

When working with 3D NumPy arrays, savetxt() and loadtxt() functions cannot directly handle them since they expect 2D arrays. To save and load 3D arrays, you need to reshape them to 2D format first, then reshape back to 3D after loading.

The Problem with 3D Arrays

Using savetxt() or loadtxt() with 3D arrays directly throws an error:

ValueError: Expected 1D or 2D array, got 3D array instead

Solution: Reshape Before Saving and After Loading

The solution involves three steps:

1. Reshape 3D array to 2D before saving
2. Save/load using savetxt()/loadtxt()
3. Reshape back to original 3D shape after loading

Example 1: Working with TXT Files

import numpy as np

# Create a 3D array of shape (3, 2, 2)
array_3d = np.array([[[3, 18], [46, 79]], 
                     [[89, 91], [66, 75]], 
                     [[77, 34], [21, 19]]])

print("Original 3D array:")
print(array_3d)
print("Shape:", array_3d.shape)

# Reshape to 2D for saving
array_2d = array_3d.reshape(array_3d.shape[0], -1)
print("\nReshaped 2D array:")
print(array_2d)

# Save to TXT file
np.savetxt("data.txt", array_2d)

# Load from TXT file
loaded_array = np.loadtxt("data.txt")
print("\nLoaded array shape:", loaded_array.shape)

# Reshape back to original 3D shape
restored_3d = loaded_array.reshape(array_3d.shape)
print("Restored 3D array shape:", restored_3d.shape)

# Verify arrays are identical
print("Arrays are identical:", np.array_equal(array_3d, restored_3d))

Original 3D array:
[[[ 3 18]
  [46 79]]

 [[89 91]
  [66 75]]

 [[77 34]
  [21 19]]]
Shape: (3, 2, 2)

Reshaped 2D array:
[[ 3 18 46 79]
 [89 91 66 75]
 [77 34 21 19]]

Loaded array shape: (3, 4)
Restored 3D array shape: (3, 2, 2)
Arrays are identical: True

Example 2: Working with CSV Files

import numpy as np

# Create a 3D array
array_3d = np.array([[[3, 18], [46, 79]], 
                     [[89, 91], [66, 75]], 
                     [[77, 34], [21, 19]]])

print("Original 3D array shape:", array_3d.shape)

# Reshape to 2D
array_2d = array_3d.reshape(array_3d.shape[0], -1)

# Save to CSV file with comma delimiter
np.savetxt("data.csv", array_2d, delimiter=",", fmt="%d")

# Load from CSV file
loaded_csv = np.loadtxt("data.csv", delimiter=",").astype(int)
print("Loaded CSV shape:", loaded_csv.shape)

# Reshape back to 3D
restored_3d = loaded_csv.reshape(array_3d.shape)
print("Restored shape:", restored_3d.shape)

# Verify data integrity
print("Data preserved correctly:", np.array_equal(array_3d, restored_3d))

Original 3D array shape: (3, 2, 2)
Loaded CSV shape: (3, 4)
Restored shape: (3, 2, 2)
Data preserved correctly: True

Key Points

Best Practices

• Always store the original shape before reshaping
• Use array.reshape(shape[0], -1) to flatten while preserving first dimension
• Verify data integrity with np.array_equal() after loading
• Consider using np.save()/np.load() for better performance with large 3D arrays

Conclusion

To save 3D NumPy arrays using savetxt(), reshape them to 2D first, then reshape back to 3D after loading with loadtxt(). This approach works for both TXT and CSV files while preserving data integrity.