How to set figure size in Matplotlib

A chart prepared for a slide, report, or image slot often has a fixed pixel rectangle to fill. Working backward from that output requirement gives a Matplotlib figure a deliberate canvas instead of relying on the default size.

Matplotlib interprets figsize as width and height in inches, while DPI states how many raster pixels occupy each inch. Dividing a 900 by 480 pixel target by 150 DPI produces the required 6 by 3.2 inch figure.

Setting figsize, dpi, and layout=“constrained” when the figure is created gives the layout engine the final canvas while it positions titles and labels. Saving with the default bounding box preserves that canvas; bbox_inches=“tight” recrops the content and can change the exported pixel dimensions.

1. Create figure_size.py with the 900 by 480 pixel requirement and its conversion at 150 DPI.

   figure_size.py

   from pathlib import Path
    
   import matplotlib.image as mpimg
   import matplotlib.pyplot as plt
    
    
   target_pixels = (900, 480)
   dpi = 150
   figsize = tuple(pixels / dpi for pixels in target_pixels)
   output = Path("sales-figure-size.png")

   The conversion yields a figsize of 6 by 3.2 inches. Matching figure and export DPI keeps the requested pixel dimensions predictable.

1. Add the plotted data, labeled axes, and constrained-layout figure to figure_size.py.

   quarters = ["Q1", "Q2", "Q3", "Q4"]
   revenue = [42, 51, 58, 64]
   fig, ax = plt.subplots(
       figsize=figsize,
       dpi=dpi,
       layout="constrained",
   )
    
   ax.plot(quarters, revenue, marker="o", linewidth=2.0)
   ax.set_title("Quarterly revenue")
   ax.set_xlabel("Quarter")
   ax.set_ylabel("Revenue (USD thousands)")
   ax.grid(True, axis="y", alpha=0.25)

   layout=“constrained” adjusts the axes inside the existing figure rather than enlarging the 900 by 480 pixel canvas.

2. Finish figure_size.py with a default-bounds PNG export and a saved-file dimension assertion.

   fig.savefig(output)
    
   saved_image = mpimg.imread(output)
   actual_pixels = (saved_image.shape[1], saved_image.shape[0])
   figure_width, figure_height = fig.get_size_inches()
   plt.close(fig)
    
   print(f"requested pixels: {target_pixels[0]} x {target_pixels[1]}")
   print(f"figure size: {figure_width:.2f} x {figure_height:.2f} in at {dpi} DPI")
   print(f"saved pixels: {actual_pixels[0]} x {actual_pixels[1]}")
    
   if actual_pixels != target_pixels:
       raise RuntimeError(f"expected {target_pixels}, saved {actual_pixels}")
    
   print(f"saved: {output}")

   A bbox_inches=“tight” export replaces the full figure canvas with a content-based bounding box, so its pixel dimensions can differ from the requested canvas.

1. Run figure_size.py to produce the 900 by 480 pixel PNG and its dimension report.

   $ python3 figure_size.py
   requested pixels: 900 x 480
   figure size: 6.00 x 3.20 in at 150 DPI
   saved pixels: 900 x 480
   saved: sales-figure-size.png