pose

Body pose recognition has applications in:

• image compression when only the position of the subject is needed
• social signaling
• ADAS driver monitoring system (DMS)

The human body pose consists of:

• body detection in the image
• body segmentation from the image
• mapping the 2D image to the 3D model
• tracking the pose over time (movie)

photo credit: posetrack.net

REFERENCES

• https://posetrack.net/workshops/iccv2017/
• http://vision.imar.ro/human3.6m/challenge_open.php
• https://competitions.codalab.org/competitions/19650
• https://github.com/facebookresearch/DensePose/blob/master/detectron/datasets/densepose_cocoeval.py
• 
  

Whimsical Mind Graph - Explore Machine Learning

As I am studying Machine Learning I am creating the mind graph à la Daniel Bourke:

https://whimsical.com/K85bjCAHxKg4FZBf3YW7en

Julia Flux Meatalhead.jl

## Uncomment the lines below if you have not run it previously
# import Pkg; Pkg.add("Metalhead")
# import Pkg; Pkg.add("QuartzImageIO")

using Metalhead
using Metalhead: classify
vgg = VGG19()

## local file in this lab's folder
image = load("skunk001.jpeg")
classify(vgg, image)

output:

"skunk, polecat, wood pussy"

Conda nodejs

$ conda install -c conda-forge nodejs

Julia Flux

import Pkg; Pkg.add("Flux")
using Flux

Julia CUDA

import Pkg
Pkg.add("CUDAapi")

Updating registry at `~/.julia/registries/General`
Updating git-repo `https://github.com/JuliaRegistries/General.git`
Resolving package versions... Installed CUDAapi ─ v4.0.0 Updating `~/.julia/environments/v1.4/Project.toml` [3895d2a7] + CUDAapi v4.0.0 Updating `~/.julia/environments/v1.4/Manifest.toml` [3895d2a7] + CUDAapi v4.0.0

using CUDAapi
has_CUDA = "false"
if has_cuda()
    @info "CUDA is on"
    has_CUDA = "true"
end
println("it is ", has_CUDA, " that I have CUDA.")

┌ Info: Precompiling CUDAapi [3895d2a7-ec45-59b8-82bb-cfc6a382f9b3]
└ @ Base loading.jl:1260

it is false that I have CUDA.