This repo implements $1 Unistroke Recognizer, a fast and simple 1-stroke gesture recognizer. As demonstrated in the linked image, it could identify various shapes with 10 samples of each. Adding a new gesture also doesn't require full recomputation:
This library implemented the Protractor optimization version of the original algorithm. The official gesture logs are included as part of the test data.
This library requires C++17 or higher C++ compiler.
If you use CMake, it's recommended to use CPM to add this project as a dependency:
# Replace 0.1 with the latest git vX tag
CPMAddPackage("gh:htfy96/dollar#0.1")
target_link_libraries(your-program dollar)
Alternatively, you can copy all files under include and src and manually compile them as a library.
#include "dollar.hpp"
dollar::Stroke stroke1{{{0., 0.}, {0., 1.}, {1., 1.}}, dollar::Orientation::Sensitive};
dollar::Stroke stroke2{{{0., 0.}, {1., 1.}, {1., 0.}}, dollar::Orientation::Sensitive};
vector<dollar::Stroke> samples{stroke1, stroke2};
dollar::Stroke testStroke{{{0., 0.}, {0., 0.9}, {0.02, 0.91}, {1., 1.}}, dollar::Orientation::Sensitive};
auto [it, score] = dollar::recognize(testStroke, samples.begin(), samples.end());
assert(it == samples.begin()); // returns stroke1- /~https://github.com/olwal/microdollar : microdollar provided a C-alike interface, while this project provides a modern interface
Apache 2.0
- Jacob O. Wobbrock, University of Washington
- Andrew D. Wilson, Microsoft Research
- Yang Li, University of Washington†