search

JuliaImages / meta

Place for discussions about the general structure of JuliaImages
0 stars 1 forks source link

Arijit Kar's Object Tracking GSOC Project 2018 #6

Open zygmuntszpak opened 6 years ago

zygmuntszpak commented 6 years ago

Hi all,

I'd like to warmly welcome our 2018 GSOC student Arijit (@arijitkar98). He will be working on implementing a variety of functions for object tracking. I wanted to start this thread so that we could begin discussing the API, and how the various functions that he will write will fit into the JuliaImages package.

Broadly speaking, the aim of the GSOC project will be to implement the following capabilities:

  1. Optical Flow Algorithms 1.1. Lucas-Kanade method 1.2. Farneback Dense Optical Flow method
  2. Tracking Algorithms 2.1. Boosting Tracking 2.2. Multiple Instance Learning 2.3. MedianFlow Tracking 2.4. Tracking Learning Detection framework

Should these functions live in a new ImageTracking package?

I'd also like for ask for your help in shaping the API and the overall design of these functions to ensure that they follow the best-practice.

Arijit has already put some thought into the API in his proposal. Arijit, could you please post your intentions for the Optical Flow work so that others could help shape the design of the API?

arijitkar98 commented 6 years ago

Thank you @zygmuntszpak for creating this thread.

For the optical flow algorithms, I plan to have a single method optical_flow(prevImg, nextImg, algo::OpticalFlowAlgo) and use dispatch to select the different algorithms.

LK{} and Farneback{} will be the two subtypes of OpticalFlowAlgo.