Skip to main content
W&B Fully Connected
>
Articles
>
Beginner

Visualize Model Predictions with Weights & Biases

This article explores how to visualize a model's predictions using Weights & Biases, including images, videos, audio, tables, HTML, and more.
Lavanya Shukla
Created on February 18|Last edited on November 3
In this article, we'll look at how to visualize a model's predictions with Weights & Biases – images, videos, audio, tables, HTML, metrics, plots, 3d objects and point clouds.
If you have any questions, we'd love to answer them.

Table of Contents

MetricsPlotsImagesVideosAudio3D ObjectsPoint CloudsHTMLIncremental Logging


Metrics

Calling wandb.log(dict) logs the keys and values of the dictionary passed in and associates the values with a step.

Example

# Initialize run 

wandb.init()

# Log multiple metrics 

wandb.log({'accuracy':0.9,'epoch':5})

# Log metrics in a loop

for price in apple ['close']: 
wandb.log({"Stock Price":price})


Run set
1


Plots

You can pass a matplotlib pyplot or figure object into wandb.log
By default we'll convert the plot into a plotly plot. If you want to explicitly log the plot as an image, you can pass the plot into wandb.Image. We also accept directly logging plotly charts.

Example

import matplotlib.pyplot as plt 

fibonacci = [0, 1, 1, 2, 3, 5, 8, 13, 21, 34]

plt.plot(fibonacci) 

plt.ylabel('some interesting numbers') 

# Initialize run
wandb.init()

# Log plot object

wandb.log({"plot":plot})


Run set
1


Images

If the image is a NumPy array, we assume it's grayscale if the last dimension is 1, RGB if it's 3, and RGBA if it's 4.
If the array contains floats we convert them to ints between 0 and 255. You can specify a mode manually or just supply a PIL.Image. It's recommended to log fewer than 50 images per step. Click "Create Visualization" to customize the image gallery.

Example

from matplotlib import pyplot as plt 
path_to_img = "../input/kernel-files/cafe.jpg"  
im = plt.imread(path_to_img) 

# Initialize run
wandb.init()

# Log image(s) 
wandb.log({"img":[wandb.Image(im,caption="Cafe")]})


Run set
1


Videos

If the video is a NumPy array, we assume the dimensions are: time, channels, width, and height. By default, we create a 4 fps gif image (ffmpeg and the moviepy python library is required when passing NumPy objects).
Supported formats are "gif", "mp4", "webm", and "ogg". If you pass a string to wandb.Video we assert the file exists and is a supported format before uploading to Weights & Biases. Passing a BytesIO object will create a tempfile with the specified format as the extension. On the W&B runs page, you will see your videos in the Media section.

Example

path_to_video = "../input/kernel-files/openai-gym.mp4

# Initialize run
wandb.init()

# Log video(s) 
wandb.log({"video":wandb.Video(path_to_video, fps=4, format="gif")})


Run set
1


Audio

The maximum number of audio clips that can be logged per step is 100.

Example

from scipy.io import wavfile
path_to_audio = "../input/kernel-files/piano.wav" 

# Initialize run 
wandb.init()

# Log audio file
wandb.log({"examples":[wandb.Audio(path_to_audio, caption="Piano", sample_rate=32)]})



Run set
1


3D Objects

We support logging 3D file types in three different formats: `glTF, glb, obj`. The 3D files will be viewable on the run page upon completion of your run.

Example

```python
path_to_obj = "../input/kernel-files/wolf.obj"

# Initialize run
wandb.init()

# Log 3D objective
wandb.log({"3d_object":wandb.Object3D(open(path_to_obj))})


Run set
1


Point Clouds

Point Clouds logging has currently has two modes.
Logging a single set of points representing an object , useful for representing datasets like ShapeNet(Example Report).
Along with a new beta release LiDAR scene renderer.
Logging a set of points is as simple as passing in a NumPy array containing your coordinates and the desired colors for the points.
Three different shapes of NumPy arrays are supported for flexible color schemes, supporting common ML us
  • [[x, y, z], ...] nx3
  • [[x, y, z, c], ...] nx4 | cis a category in the range [1, 14] (Useful for segmentation)
  • [[x, y, z, r, g, b], ...] nx6 | r,g,b are values in the range [0,255]for red, green, and blue color channels.

Example

# Initialize run
wandb.init()

# Log points and boxes  
wandb.log(
        {
            "point_scene": wandb.Object3D(
                {
                    "type": "lidar/beta",
                    "points": np.array(
                        [
                            [0.4, 1, 1.3], 
                            [1, 1, 1], 
                            [1.2, 1, 1.2]
                        ]
                    ),
                    "boxes": np.array(
                        [
                            {
                                "corners": [
                                    [0,0,0],
                                    [0,1,0],
                                    [0,0,1],
                                    [1,0,0],
                                    [1,1,0],
                                    [0,1,1],
                                    [1,0,1],
                                    [1,1,1]
                                ],
                                "label": "Box",
                                "color": [123,321,111],
                            },
                            {
                                "corners": [
                                    [0,0,0],
                                    [0,2,0],
                                    [0,0,2],
                                    [2,0,0],
                                    [2,2,0],
                                    [0,2,2],
                                    [2,0,2],
                                    [2,2,2]
                                ],
                                "label": "Box-2",
                                "color": [111,321,0],
                            }
                        ]
                    ),
                }
            )
        }
    )


Run set
1


HTML

Custom HTML can be logged at any key, this exposes an HTML panel on the run page. By default we inject default styles, you can disable default styles by passing inject=False

Example

path_to_html = "../input/kernel-files/some_html.html" 

# Initialize run
wandb.init()

# Log plot HTML
# As a file
wandb.log({"custom_file": wandb.Html(open(path_to_html))})

# Inline
wandb.log({"custom_string": wandb.Html('<a href="https://mysite">Link</a>')})

# Disable default styles
wandb.log({"custom_file": wandb.Html(open("some.html"), inject=False)})


Run set
1


Incremental Logging

If you want to log to a single history step from lots of different places in your code you can pass a step-index to wandb.log() as shown in the example below.
wandb.log(dict) accepts a few keyword arguments:
  • step — Step to associate the log with (see Incremental Logging)
  • commit — If true, increments the step associated with the log(default: true)

Example

# Initialize run
wandb.init()

# Log the metric to specific steps
for step, price in enumerate(apple['close']):
    # Specify steps
    wandb.log({"Stock Price": price}, step=step)

# Initialize a new run
wandb.init(project="visualize-metrics", name="incremental_logging")

# Log many metrics to the same step
for close, date in zip(apple['close'], apple['date']):
    # Log multiple metrics to the same step
    wandb.log({"Stock Price": close}, commit=False)
    
    # Somewhere else when I'm ready to report this step:
    wandb.log({'Another Metric': close*10})


Run set
2

As long as you keep passing the same value for step, W&B will collect the keys and values from each call in one unified dictionary. As soon you call wandb.log() with a different value for step than the previous one, W&B will write all the collected keys and values to the history, and start collection over again.
Note that this means you should only use this with consecutive values for step: 0, 1, 2, .... This feature doesn't let you write to absolutely any history step that you'd like, only the "current" one and the "next" one.
You can also set commit=False in wandb.log to accumulate metrics, just be sure to call wandb.log without the commit flag to persist the metrics.
Thar Htet San
Thar Htet San •  
Thank for sharing
Reply
Milad Aghajohari
Milad Aghajohari •  
Thanks for explaining how to log matplotlib plots.
Reply
Hongbo Miao
Hongbo Miao •  
This looks so cool! Thanks for sharing!
Reply
Tags: Beginner, Domain Agnostic, Tutorial, W&B Meta, Plots

Created with ❤️ on Weights & Biases.

https://wandb.ai/lavanyashukla/visualize-predictions/reports/Visualize-Model-Predictions-with-Weights-Biases--Vmlldzo1NjM4OA
Made with Weights & Biases. Sign up or log in to create reports like this one.
Iterate on AI agents and models faster. Try Weights & Biases today.
Sign upTry W&B now