PyEXR

A simple EXR IO-library for Python that simplifies the use of OpenEXR.

Installation

pip install pyexr

Reading

Simple files

import pyexr

with pyexr.open("color.exr") as file:
    file.channels               # [R, G, B]
    file.width                  # 1280
    file.height                 # 720
    file.channel_precision["R"] # pyexr.FLOAT

    img = file.get()                         # (720,1280,3) np.float32 array
    img = file.get(precision=pyexr.HALF)          # (720,1280,3) np.float16 array
    red = file.get("R")                  # (720,1280,1) np.float32 array
    red = file.get("R", precision=pyexr.HALF) # (720,1280,1) np.float16 array

Fat / Multi-channel EXRs

import pyexr

with pyexr.open("multi-channel.exr") as file:
    file.channels               # [R, G, B, A, Variance.R, Variance.G, Variance.B]
    file.width                  # 1280
    file.height                 # 720

    all = file.get()            # (720,1280,7) np.float32 array (R,G,B,A,Var..)
    var = file.get("Variance")  # (720,1280,3) np.float32 array
    col = file.get("default")   # (720,1280,4) np.float32 array (R,G,B,A)
    file.channel_map['default'] # ['R','G','B','A']

    var_r = file.channel("Variance.R") # (720,1280,3) np.float32 array

One line reading

import pyexr

# 'color.exr' contains R, G, B
img = pyexr.read("color.exr")                  # (720,1280,3) np.float32 array
img = pyexr.read("color.exr", precision=pyexr.HALF) # (720,1280,3) np.float16 array

# 'multi-channel.exr' contains R, G, B, A, Variance.R, Variance.G, Variance.B
all = pyexr.read("multi-channel.exr")             # (720,1280,7) np array
col = pyexr.read("multi-channel.exr", "default")  # (720,1280,4) np array
var = pyexr.read("multi-channel.exr", "Variance") # (720,1280,3) np array

col, var = pyexr.read("multi-channel.exr", ["default", "Variance"])
col, var = pyexr.read("multi-channel.exr", ["default", "Variance"], precision=pyexr.HALF)
col, var = pyexr.read("multi-channel.exr", ["default", "Variance"], precision=[pyexr.HALF, pyexr.FLOAT])

Writing

You can write a matrix to EXR without specifying channel names. Default channel names will then be used:

# columns	default
1 channel	Z
2 channels	X, Y
3 channels	R, G, B
4 channels	R, G, B, A

import pyexr

depth  # (720,1280) np.float16 array
color  # (720,180,3) np.float32 array
normal # (720,180,3) np.float32 array

pyexr.write("out.exr", depth)            # one FLOAT channel: Z
pyexr.write("out.exr", color)            # three FLOAT channels: R, G, B
pyexr.write("out.exr", normal,
          channel_names=['X','Y','Z']) # three FLOAT channels: X, Y, Z

Writing Fat EXRs

import pyexr

depth    # (720,1280) np.float16 array
color    # (720,180,3) np.float32 array
variance # (720,180,3) np.float32 array

data = {'default': color, 'Depth': depth, 'Variance': variance} # default is a reserved name

pyexr.write("out.exr", data) # channels R, G, B, Depth.Z, Variance.(R,G,B)

# Full customization:
pyexr.write(
    "out.exr",
    data,
    precision = {'default': pyexr.HALF},
    channel_names = {'Depth': ['Q']}
) # channels R, G, B, Depth.Q, Variance.R, Variance.G, Variance.B