Depth & Column Stacking (dstack, column_stack, stack)0%

Splitting Arrays into Chunks (split, hsplit, vsplit)

Depth & Column Stacking (dstack, column_stack, stack)

Beginner12 min read•Updated: Jul 11, 2026

Depth & Column Stacking: dstack(), column_stack(), & stack()

While vstack and hstack handle basic 2D horizontal and vertical stacking, real-world data science and computer vision problems frequently require working along other axes:

Merging independent grayscale channels into a multi-channel RGB or RGBA image tensor.
Stacking 1D feature vectors side-by-side as columns in a 2D design matrix.
Stacking arrays along an entirely new dimension (such as stacking multiple time-series frames).

1. Column Stacking: `np.column_stack()`

np.column_stack() takes a sequence of 1D arrays and stacks them as columns into a single 2D matrix. This is one of the cleanest, most idiomatic ways in NumPy to assemble tabular feature matrices from separate 1D variables:

1import numpy as np
2 
3# Feature vectors
4feature_age = np.array([25, 30, 45, 52])
5feature_income = np.array([55000, 68000, 110000, 95000])
6feature_credit = np.array([710, 680, 790, 740])
7 
8# Assemble into a 2D design matrix (4 samples x 3 features)
9design_matrix = np.column_stack((feature_age, feature_income, feature_credit))
10print("Design Matrix (Shape:", design_matrix.shape, "):
11", design_matrix)

Output:

1Design Matrix (Shape: (4, 3) ):
[[    25  55000    710]
[    30  68000    680]
[    45 110000    790]
[    52  95000    740]]

Note

For 2D arrays, column_stack() behaves identically to hstack().

2. Depth Stacking: `np.dstack()`

np.dstack() stacks arrays along axis 2 (the depth axis).

This is the standard operation for assembling color images from separate 2D color channel planes:

1# Create three 2D color channels (e.g. 4x4 image)
2red_channel   = np.full((4, 4), 255, dtype=np.uint8)
3green_channel = np.full((4, 4), 128, dtype=np.uint8)
4blue_channel  = np.zeros((4, 4), dtype=np.uint8)
5 
6# Stack along depth to construct an RGB image tensor of shape (4, 4, 3)
7rgb_image = np.dstack((red_channel, green_channel, blue_channel))
8print("RGB Image Tensor shape:", rgb_image.shape)  # (4, 4, 3)
9print("Pixel (0, 0) RGB values:", rgb_image[0, 0])  # [255, 128, 0]

3. Creating New Axes with `np.stack()`

Notice the critical conceptual difference between np.concatenate() and np.stack():

np.concatenate(): Joins arrays along an existing axis. Output has the same number of dimensions (ndim).
np.stack(): Joins arrays along a NEW axis. Output has ndim + 1 dimensions.

1frame1 = np.ones((10, 10))
2frame2 = np.ones((10, 10)) * 2
3frame3 = np.ones((10, 10)) * 3
4 
5# Stack 2D frames into a 3D video clip along axis 0: shape (3, 10, 10)
6video_clip = np.stack((frame1, frame2, frame3), axis=0)
7print("Stacked axis 0 shape:", video_clip.shape)
8 
9# Stack along axis -1: shape (10, 10, 3)
10video_channels = np.stack((frame1, frame2, frame3), axis=-1)
11print("Stacked axis -1 shape:", video_channels.shape)

4. Summary of Stacking Helpers

Function	Effective Behavior	Dimensionality Change	Typical Use Case
`np.concatenate`	Joins along existing axis	Same `ndim`	Appending rows or columns
`np.stack`	Joins along brand new axis	`ndim + 1`	Batching images, time-series frames
`np.vstack`	Row-wise stacking	Promotes 1D $\to$ 2D	Appending sample rows
`np.hstack`	Column-wise for 2D, 1D concat	Same or 2D	Horizontal image tiling
`np.column_stack`	1D arrays as columns	Always 2D	ML design matrices
`np.dstack`	Stacks along 3rd axis (depth)	Promotes to 3D	RGB channel merging

Interactive Quiz• Knowledge Check

Multiple Choice Questions

Question 1

How does np.stack() differ fundamentally from np.concatenate()?

Question 2

If you have three 1D arrays of shape (100,), what is the shape of the output from np.column_stack((a, b, c))?

Question 3

Which function is most appropriate for combining three 2D matrices of shape (1080, 1920) representing Red, Green, and Blue into a standard image array of shape (1080, 1920, 3)?

Question 4

What is the output shape of np.stack([np.zeros((4, 5)), np.zeros((4, 5))], axis=1)?

Question 5

When using np.stack(), what condition must all input arrays satisfy?

Next Lesson

Splitting Arrays into Chunks (split, hsplit, vsplit)

Continue learning with hands-on practice, examples, and exercises in the upcoming topic.

Related Lessons

Previous Lesson	Next Lesson
Concatenating & Stacking Arrays (concatenate, vstack, hstack)	Splitting Arrays into Chunks (split, hsplit, vsplit)

Practice Quiz

Test your understanding of this lesson with 5 questions. Each question has one correct answer.

Previous Next

PreviousPrevNext

MSK Institute • Shikohabad

Loading Page Content...

Preparing courses, live labs, and learning resources.सामग्री लोड हो रही है, कृपया प्रतीक्षा करें...

Connecting serverPlease wait...

💻 Practical Labs🚀 Free Demo📜 Verified Certificates

Splitting Arrays into Chunks (split, hsplit, vsplit)

Depth & Column Stacking (dstack, column_stack, stack)

Beginner12 min read•Updated: Jul 11, 2026

Study Materials Cheatsheet

Depth & Column Stacking: dstack(), column_stack(), & stack()

While vstack and hstack handle basic 2D horizontal and vertical stacking, real-world data science and computer vision problems frequently require working along other axes:

Merging independent grayscale channels into a multi-channel RGB or RGBA image tensor.
Stacking 1D feature vectors side-by-side as columns in a 2D design matrix.
Stacking arrays along an entirely new dimension (such as stacking multiple time-series frames).

1. Column Stacking: `np.column_stack()`

1import numpy as np
2 
3# Feature vectors
4feature_age = np.array([25, 30, 45, 52])
5feature_income = np.array([55000, 68000, 110000, 95000])
6feature_credit = np.array([710, 680, 790, 740])
7 
8# Assemble into a 2D design matrix (4 samples x 3 features)
9design_matrix = np.column_stack((feature_age, feature_income, feature_credit))
10print("Design Matrix (Shape:", design_matrix.shape, "):
11", design_matrix)

Output:

1Design Matrix (Shape: (4, 3) ):
[[    25  55000    710]
[    30  68000    680]
[    45 110000    790]
[    52  95000    740]]

Note

For 2D arrays, column_stack() behaves identically to hstack().

2. Depth Stacking: `np.dstack()`

np.dstack() stacks arrays along axis 2 (the depth axis).

This is the standard operation for assembling color images from separate 2D color channel planes:

1# Create three 2D color channels (e.g. 4x4 image)
2red_channel   = np.full((4, 4), 255, dtype=np.uint8)
3green_channel = np.full((4, 4), 128, dtype=np.uint8)
4blue_channel  = np.zeros((4, 4), dtype=np.uint8)
5 
6# Stack along depth to construct an RGB image tensor of shape (4, 4, 3)
7rgb_image = np.dstack((red_channel, green_channel, blue_channel))
8print("RGB Image Tensor shape:", rgb_image.shape)  # (4, 4, 3)
9print("Pixel (0, 0) RGB values:", rgb_image[0, 0])  # [255, 128, 0]

3. Creating New Axes with `np.stack()`

Notice the critical conceptual difference between np.concatenate() and np.stack():

np.concatenate(): Joins arrays along an existing axis. Output has the same number of dimensions (ndim).
np.stack(): Joins arrays along a NEW axis. Output has ndim + 1 dimensions.

1frame1 = np.ones((10, 10))
2frame2 = np.ones((10, 10)) * 2
3frame3 = np.ones((10, 10)) * 3
4 
5# Stack 2D frames into a 3D video clip along axis 0: shape (3, 10, 10)
6video_clip = np.stack((frame1, frame2, frame3), axis=0)
7print("Stacked axis 0 shape:", video_clip.shape)
8 
9# Stack along axis -1: shape (10, 10, 3)
10video_channels = np.stack((frame1, frame2, frame3), axis=-1)
11print("Stacked axis -1 shape:", video_channels.shape)

4. Summary of Stacking Helpers

Function	Effective Behavior	Dimensionality Change	Typical Use Case
`np.concatenate`	Joins along existing axis	Same `ndim`	Appending rows or columns
`np.stack`	Joins along brand new axis	`ndim + 1`	Batching images, time-series frames
`np.vstack`	Row-wise stacking	Promotes 1D $\to$ 2D	Appending sample rows
`np.hstack`	Column-wise for 2D, 1D concat	Same or 2D	Horizontal image tiling
`np.column_stack`	1D arrays as columns	Always 2D	ML design matrices
`np.dstack`	Stacks along 3rd axis (depth)	Promotes to 3D	RGB channel merging

Interactive Quiz• Knowledge Check

Multiple Choice Questions

Question 1

How does np.stack() differ fundamentally from np.concatenate()?

Question 2

If you have three 1D arrays of shape (100,), what is the shape of the output from np.column_stack((a, b, c))?

Question 3

Which function is most appropriate for combining three 2D matrices of shape (1080, 1920) representing Red, Green, and Blue into a standard image array of shape (1080, 1920, 3)?

Question 4

What is the output shape of np.stack([np.zeros((4, 5)), np.zeros((4, 5))], axis=1)?

Question 5

When using np.stack(), what condition must all input arrays satisfy?

Next Lesson

Splitting Arrays into Chunks (split, hsplit, vsplit)

Continue learning with hands-on practice, examples, and exercises in the upcoming topic.

Related Lessons

Previous Lesson	Next Lesson
Concatenating & Stacking Arrays (concatenate, vstack, hstack)	Splitting Arrays into Chunks (split, hsplit, vsplit)

Practice Quiz

Test your understanding of this lesson with 5 questions. Each question has one correct answer.

Previous Next

Depth & Column Stacking: dstack(), column_stack(), & stack()

1. Column Stacking: np.column_stack()

2. Depth Stacking: np.dstack()

3. Creating New Axes with np.stack()

4. Summary of Stacking Helpers

Multiple Choice Questions

Splitting Arrays into Chunks (split, hsplit, vsplit)

Related Lessons

Practice Quiz

Loading Page Content...

Depth & Column Stacking: dstack(), column_stack(), & stack()

1. Column Stacking: np.column_stack()

2. Depth Stacking: np.dstack()

3. Creating New Axes with np.stack()

4. Summary of Stacking Helpers

Multiple Choice Questions

Splitting Arrays into Chunks (split, hsplit, vsplit)

Related Lessons

Practice Quiz

1. Column Stacking: `np.column_stack()`

2. Depth Stacking: `np.dstack()`

3. Creating New Axes with `np.stack()`

1. Column Stacking: `np.column_stack()`

2. Depth Stacking: `np.dstack()`

3. Creating New Axes with `np.stack()`