matplotlib, savefig: DPI setting is ignored

Question

I cannot find anyone else with this problem. In matplotlib, you can view your plots using either show() or savefig(). These generate slightly different images; in my case, the savefig() image is uglier and harder to understand. I need to make life easy for my examinator, so..

I found some topics that suggested I set the DPI size to match that of show(). I have tried:

-> Setting savefig.dpi directly with matplotlib.rcParams['savefig.dpi'] = 80.

-> Setting savefig.dpi directly in ~/.matplotlib/matplotlibrc.

-> Moving my rc file to CWD.

-> Finally, using savefig('image.pdf', dpi=80)

I can verify that the attribute is indeed getting set; but it seems that the setting is igored by savefig(). Can anyone help with this?

(Simplified) code:

plt.bar(ind, functimes, yerr=stddev, bottom=0, width=barWidth, align='center', color='b')

ax = plt.gca()
ax.legend(barRcts, barLegend)
plt.title("Function Call Overhead")
plt.xlabel("Function ID")
plt.ylabel("Execution Time [us]")

plt.xticks(ind, funcNames)
figtest.autofmt_xdate()

plt.savefig(out_file, dpi=80, format='pdf')

Answer 1

PDF saving uses the DPI setting for image resolution inside the PDF, not for any line/polygon resolution (which is meaningless in a vector format).

If your happy with the output on screen, and don't need to have a scalable graphic, saving as png is probably fine.

Answer 2

On your IDE, when you pass the cursor over the function plt.savefig() it shows the next: (function) savefig: (*args: Any, **kwargs: Any) it means that the function only accepts unlimited arguments (*args) and Keyworded Arguments (**kwargs). So, if you go the documentation of Matplotlib you will find the arguments you could use https://matplotlib.org/3.1.1/api/_as_gen/matplotlib.pyplot.savefig.html.

Now, if you only want to present a good quality plots you could do the next as an example. Open a jupyter notebook on vscode and code plt.savefig("my_image.jpg", dpi = 300) [dpi stands for dots per inch, if you use a higher dpi you will get a high quality image] then with markdown show your plots with the following code:

plt.plot(a,v_t_retraso, 'r', label = 'FP - retraso')
plt.plot(a,v_t_adelanto,'b', label = 'FP - adelanto')

plt.title('Caracteristicas de los terminales', fontdict={'fontname': 'Comic Sans MS', 'fontsize': 20})

# Escalando
plt.yticks([400,450,500,550])
plt.grid(True)
plt.legend()
plt.xlabel('Corriente en Linea, $A$')
plt.ylabel('Voltaje en los terminales, $V$')
plt.savefig('terminales_4.jpg', dpi=300)
plt.close()

plt.plot(v_t_retraso,a)
plt.xlabel('Corriente en Linea, $A$')
plt.ylabel('Voltaje en los terminales, $V$')
plt.savefig('terminales_5.jpg', dpi=300)
plt.close()

With the save images you could open a cell on markdown and write the following code:

 <!-- If you want a full-size image then use the following code: -->
 <!-- ![](terminales_4.jpg)  ![](terminales_5.jpg) -->
 <!-- For custom size -->
<img src="terminales_4.jpg" alt="terminales_4" style="width:432px;"/>
<img src="terminales_5.jpg" alt="terminales_5" style="width:432px;"/>

 

Finally you could export your code as HTML or Pdf and show as many graphs as you want. An example is shown in the next image.