Using the Visual Debugger - nanpuhaha/SerpentAI GitHub Wiki

The Serpent.AI framework ships with a small desktop application called the Visual Debugger. Its role is simple: Help the developer visualize in-memory image data at game agent runtime. To make an analogy, the Visual Debugger is like a print statement but for images!

Launching the Visual Debugger

Like most things in the framework, launching the Visual Debugger is done through a serpent command.

Run serpent visual_debugger

You should see an application looking similar to this open up:

Sending Image Data to the Visual Debugger

A Visual Debugger instance exposes a function to store image data to be consumed by the application.

self.store_image_data

Method Signature store_image_data(self, image_data, image_shape, bucket="debug")

• image_data: A numpy ndarray of shape (rows, cols, color_channels) and dtype np.uint8
• image_shape: The shape of image_data
• bucket: Which labeled bucket to display the image in in the Visual Debugger application

Example

import skimage.io
from serpent.visual_debugger.visual_debugger import VisualDebugger

image = skimage.io.imread("image.png")

visual_debugger = VisualDebugger()
visual_debugger.store_image_data(
    image_data = image,
    image_shape = image.shape,
    bucket = "0"
)

VisualDebugger instances can be instantiated anywhere in your code. Game Agents already have an instance in scope as self.visual_debugger.

Customizing the Visual Debugger's Buckets

By default, the Visual Debugger will expect image data in the following buckets: "0", "1", "2" and "3". These buckets can be customized by editing config/config.yml. The relevant snippet looks like this:

visual_debugger:
    redis_key_prefix: SERPENT:VISUAL_DEBUGGER
    available_buckets:
        - "0"
        - "1"
        - "2"
        - "3"

Simply replace available_buckets with different keys. The Visual Debugger will adapt its display and allocate space for every configured bucket.