-
I’m happy to announce the release of PPN.jl which is a Deep Learning based Pose Estimation application that utilizes Chainer and Flux.jl. It includes demo script that captures image from camera (via VideoIO.jl, Images.jl), runs Flux model and finally visualize result (via ImageDraw.jl, Makie.jl)with realtime speed without GPU acceleration.
- Chainer is a deep learning framework with Python and Flux.jl is, you know.
- My application is based on Idein/chainer-pose-proposal-net which is a yet another implementation of Pose Proposal Networks by Taiki SEKII (ECCV2018).
-
I’ve also implemented Gomah.jl which provides weight converter from Chainer to Flux via PyCall.jl. Therefore we can utilize Chainer model with Julia.
-
Check our my repository to learn more.
8 Likes