Face Counting

This tutorial will get you started with wasmVision by showing how to build a face counting application that saves the counts to a BoltDB database.

Detect faces

First, let us run the built-in face detector using the wasmvision run command:

wasmvision run -p facedetectyn.wasm

wasmVision should automatically download both the facedetectyn.wasm processor and also the yunet_2023mar computer vision model needed for detecting faces.

Point your browser to http://localhost:8080 and you should see the output.

Create config file

Next, let’s do the same thing as in the first step, but use a configuration file so we can more easily change the various parameters needed for our face counting application.

wasmVision supports both TOML and YAML formats for configuration files.

Create a new file named facecount.toml (or facecount.yaml) and make sure it contains the following data:

• Config:
• TOML
• YAML

[main]
logging = "warn"

[processing]
pipeline = [
  "facedetectyn.wasm"
]

main:
  logging: "warn"

processing:
  pipeline:
    - "facedetectyn.wasm"

Run using config file

Now we can run wasmVision again, but this time using the config file we just created:

• Config:
• TOML
• YAML

wasmvision run -f facecount.toml

wasmvision run -f facecount.yaml

Point your browser to http://localhost:8080 and you should see the same output.

Count detected faces

Now we will add another processor. This processor is the built-in face-counter.wasm processor that can use the face information from the facedetectyn.wasm processor to create the count of faces detected.

Change the logging to info so we can see the output on the terminal:

• Config:
• TOML
• YAML

[main]
logging = "info"

[processing]
pipeline = [
  "facedetectyn.wasm",
  "face-counter.wasm"
]

main:
  logging: "info"

processing:
  pipeline:
    - "facedetectyn.wasm"
    - "face-counter.wasm"

Run wasmVision again:

• Config:
• TOML
• YAML

wasmvision run -f facecount.toml

wasmvision run -f facecount.yaml

You should see output similar to the following including the JSON data of the face counts:

2025/04/16 11:17:50 INFO {"timestamp":"2025-04-16T09:17:50.882075Z","average-faces-seen":1}

Save data to BoltDB

Next switch to using the BoltDB database to store the face counts on disk.

Add a new datastorage key with the value "boltdb" into the main section.

Also change the logging back to warn to reduce the amount of output on the terminal:

• Config:
• TOML
• YAML

[main]
logging = "warn"
datastorage = "boltdb"

[processing]
pipeline = [
  "facedetectyn.wasm",
  "face-counter.wasm"
]

main:
  logging: "warn"
  datastoreage: "boltdb"

processing:
  pipeline:
    - "facedetectyn.wasm"
    - "face-counter.wasm"

We need to tell wasmVision where the database file is located using the WASMVISION_STORAGE_BOLTDB_FILENAME environmental variable:

export WASMVISION_STORAGE_BOLTDB_FILENAME="facedata.db"

Now run the command again:

• Config:
• TOML
• YAML

wasmvision run -f facecount.toml

wasmvision run -f facecount.yaml

The face count data should now be getting stored into the BoltDB database.

Check BoltDB data

Let’s check that the data is being stored into the BoltDB database as expected.

Install BoltDB command line utility, if you do not already have it:

go install go.etcd.io/bbolt/cmd/bbolt@latest

Now we can list the data that has been stored:

bbolt keys facedata.db face-counter

You should see output similar to this:

2025-04-16T09:29:47.294629Z
2025-04-16T09:29:57.340648Z

Use one of the keys displayed to retrieve the data:

bbolt get facedata.db face-counter 2025-04-16T09:29:47.294629Z

You should see output similar to this:

{"timestamp":"2025-04-16T09:29:47.294629Z","average-faces-seen":1}

Turn off face box drawing

Now turn off the drawing of the face boxes in the output, since it is not needed.

To do this add a new configuration section to the file, and set the detect-draw-faces key to "false".

This configuration setting is used by the facedetectyn.wasm processor to turn on or off the drawing of the detected faces on the output.

• Config:
• TOML
• YAML

[main]
logging = "warn"
datastorage = "boltdb"

[processing]
pipeline = [
  "facedetectyn.wasm",
  "face-counter.wasm"
]

[configuration]
detect-draw-faces="false"

main:
  logging: "warn"
  datastoreage: "boltdb"

processing:
  pipeline:
    - "facedetectyn.wasm"
    - "face-counter.wasm"

configuration:
  detect-draw-faces: "false"

Run the command again:

• Config:
• TOML
• YAML

wasmvision run -f facecount.toml

wasmvision run -f facecount.yaml

Point your browser to http://localhost:8080 and you should see that output no longer shows the face boxes for each face detected.

Blur faces for privacy

The last step is for us to blur the detected faces to protect the privacy of the detected individuals.

Add another built-in processor named faceblur.wasm. This processor uses the same information that the face-count.wasm processor does, but with the purpose of blurring the detected faces in the resulting output.

• Config:
• TOML
• YAML

[main]
logging = "warn"
datastorage = "boltdb"

[processing]
pipeline = [
  "facedetectyn.wasm",
  "face-counter.wasm",
  "faceblur.wasm"
]

[configuration]
detect-draw-faces="false"

main:
  logging: "warn"
  datastoreage: "boltdb"

processing:
  pipeline:
    - "facedetectyn.wasm"
    - "face-counter.wasm"
    - "faceblur.wasm"

configuration:
  detect-draw-faces: "false"

Run the command again:

• Config:
• TOML
• YAML

wasmvision run -f facecount.toml

wasmvision run -f facecount.yaml

Point your browser to http://localhost:8080 and you should see that output now blurs each face detected.