The ml5.js team is so excited to announce some new changes to the project.
We have a new release! 🎉
This release introduces two new models ported from Tensorflow.js: Handpose and Facemesh. Handpose can be used to predict the location of 21 3D landmarks on the palm and fingers of a single hand. Facemesh can be used to predict 486 3D landmarks on the human face and can work on multiple faces at once. Facemesh provides more detail about the positions of key locations on the face than any of the models currently supported in ml5.js’s implementation of FaceApi.
In addition to new models, the 0.6.0 release also includes a variety of fixes and improvements by our community of contributors and maintainers:
- Updated tf.js version to 1.7 (up from 1.2) (#846)
- Documentation for the Object Detector class
- Switched fetch calls in core library code to Axios for improved browser support (#979)
- Added nearestFromSet function to Word2Vec (#1015)
- Fixes to the Neural Network class (#919, #1096)
- Increase the performance of the FaceAPI by supporting the Tiny Face model (#936)
We’re also going to be using this Medium page to have more direct connection to our community. In addition to our editorial profiles we’ll be including casual updates on the state of the library and featuring small creative projects in round-ups. Do you have a project you’d like featured on the website or the Medium post? Submit it here!
Lastly, the team has been hard at work investigating the current landscape of machine learning and defining a set of standards that we want to uphold. These standards reflect our commitment to keeping ml5.js a friendly, safe, and accessible community for all learners. Our investigation has culminated in a living Code of Conduct and license that we’ll be releasing for feedback later this month.
What would you like to see next from the ml5.js team? We would love to hear from you! Let us know in the comments or on Twitter!