PlotlyJS docs, https://github.com/plotly/plotlyjs.jl-docs/tree/master/julia,
do not contain any example of plot involving the trace image. Tha's why we discuss here how this trace can be defined.
using PlotlyJS, Base64
The interrogation:
?PlotlyJS.image
returns two methods for this function. The first method, image(; kwargs...) is designated to display a digital image (png, jpg, tiff/tif, ...), while the second one, to display an Array(UInt8, 3) as image.
We discuss here the first method. The second one is presented in https://github.com/empet/PlotlyJS-image-tutorials/blob/main/1p-Image-from-ArrayUInt8.ipynb.
Inspecting the reference
https://plotly.com/julia/reference/image/ we notice that the most important keyword, in the first method is the image source, described as:
Parent: data[type=image]
Type: string
Specifies the data URI of the image to be visualized. The URI consists of "data:image/[<media subtype>][;base64],<data>"
JuliaImages offers many possibilities to display an image, but creating a plot
with PlotlyJS.image, as trace, allows to associate the image title, to add axes ticks and labels,
annotations, and moreover we can flip between a list(vector) of images, selected from a dropdown menu or display a sequence of images with an attached slider.
Before creating a plot from an image, we have to encode that image as a Base64 string, using a the function bs64string, with two methods:
bs64string(data::AbstractVector) = "data:image/png;base64,"*base64encode(data)
bs64string(img::AbstractMatrix)="data:image/png;base64,$(stringmime(MIME("image/png"), img))"
bs64string (generic function with 2 methods)
The first method will be called when the image is read, with Base.read("imgfile.ext"), that returns a Vector{UInt8},
while the second one, when the image is loaded as a Matrix{RGB{T}} or
a Matrix{RGBA{T}}, via img=Images.load("imgfile.ext"). The image file can have ext =png,jpg,tiff, etc. To read awebpimage, useread("imagefile.webp"), becauseImages.loaddoes not acceptwebp` files.
The first method is much more performant than the second one.
img = read("images/sun-flower.jpg")
fig1 = Plot(PlotlyJS.image(;source=bs64string(img)), Layout(width=400, height=400,
title_text="<b>Sunflower</b>", title_x=0.5, font_family="Balto"));
#add annotations and a shape (here a white vertical line)
relayout!(fig1, annotations = [attr(
font_size=16,
font_color="RoyalBlue",
showarrow=false,
text="<b>Blue ciel</b>",
x=50, y=30, #text position
xref="x",
yref="y",
xanchor="center",
yanchor="bottom",
)])
add_shape!(fig1, #add a white line
line(x0=258, y0=168,
x1=258, y1=244,
line_color="white",
line_width=4))
display(fig1)
The original image can be saved with added elements, via PlotlyJS:
#savefig(fig1, "images/annotated-image.jpeg", width=400, height=400, scale=1)
Let us try to display a webp image (as far as I know Python px.imshow cannot display this image type), and add drawing buttons to modebar. To draw rectangles for example, set in layout, dragmode="drawrect", and to change the drawing tool click on the corresponding button in modebar:
vecuint8 = read("images/test-image.webp") #https://developers.google.com/speed/webp/gallery1
fig2 = Plot(PlotlyJS.image(;source=bs64string(vecuint8)),
Layout(width=500, height=500, title_text="<b>Webp image</b>",
title_x=0.5, font_family="Balto", dragmode="drawrect"),
config = PlotConfig(modeBarButtonsToAdd=[
"drawline",
"drawopenpath",
"drawclosedpath",
"drawcircle",
"drawrect",
"eraseshape"
]))
Hence we can add shape to the image either to an exact position (fig1) or at approximate position, with the drawing tool (fig2).