facegrep

I want to learn Rust and found about Rustface (which is a port of C++ SeetaFace Engine). It detects faces in a picture. What interesting is it doesn't need external dependencies (eg. opencv).

I think it'll be a good exercise.

Goal

• Use the example code from rustface as base
• Get model path from environment variable so users don't have to put it as argument
• Show output file (image file with bounding boxes)
• Add json support

Usage

# Clone repo
git clone --depth 1 https://github.com/wzulfikar/facegrep

# Download model from Rustface repo
curl -L -O --output-dir /tmp https://github.com/atomashpolskiy/rustface/raw/master/model/seeta_fd_frontal_v1.0.bin

# Expose the model as environment variable
export RUSTFACE_MODEL=/tmp/seeta_fd_frontal_v1.0.bin

# Run the binary with the test image
cargo run assets/test/scientists.jpg

# Use --json flag to show json output
cargo run assets/test/scientists.jpg --json

If you don't want to install Rust cargo:

• for M1, try use the binary at bin/facegrep-m1
• for Windows (Rust target x86_64-pc-windows-gnu), try use bin/facegrep-win

You'll still need to download the model and expose its path as environment variable.

Preview

▲▼▲

Note to self:

• JSON should be properly handled instead of just string-concatenated (main.rs#L57, L69)
• If I want to reuse the code in web, how difficult is it to port to wasm? Would there be performance penalty?
• Took me 1 hour to put this together. How long will it take me to do similar thing after I learn more Rust?