Use your device's camera to run object detection on the world around you.
You need to install binaryen in order to have
access to wasm-opt, which will help to shirkn the final
size of the wasm binary.
Through a combination of building in release mode, compiling
with Link Time Optimizations (LTO), telling LLVM to optimize
for size instead of speed, and running wasm-opt, we can
reduce the size of the final wasm binary by as much as 80%.
We recommend reading https://rustwasm.github.io/book/reference/code-size.html#optimizing-builds-for-code-size for additional information.
See our build script for the exact commands that we use to build the application.
See the [profile.release] section of Cargo.toml to
take a look at the LLVM optimizations used during the build.
The following command will keep a copy of the app running locally, if you want to play around with the code.
cargo web start --target=wasm32-unknown-unknownMuch of the HTML and javascript basis were made possible through openly licensed work from the Tensorflow.js COCO-SSD model. Thank you.
The drawing of bounding boxes was almost entirely made possible by Paul Irish's demo of multi-touch. Thank you.