Skip to content

Diary

Jump to bottom Edit New page
dancek edited this page Dec 14, 2012 · 72 revisions

Current TODO

  • write thesis according to Schedule
  • find out how the 60fps mode could be used
  • (change chest and hip implementation for the cylinder man)
  • consider texturing

Diary

2012-12-14

  • sent v0.6 to Tassu, with almost-finished Discussion

2012-12-01

  • made remaining corrections to Approach based on Tuukka's feedback
  • sent thesis to Tassu (v0.5)

2012-11-30

  • discussed chapters 4 and 5 with Tassu
  • decided to drop Results and only write Discussion including a recap of what we achieved and what is lacking

2012-11-29

  • got comments to my literature section (v0.3) from Tuukka
  • gave my methods section (v0.4) to Tuukka to read

2012-11-19

  • upped the pdf page count today 39->44
  • quite happy with the Literature section, though three sections have TODOs

2012-11-15/16

  • writing...
  • measurement work

2012-11-14

  • got Klaus's stuff to work for me
  • ported SkeletonPractice to Eclipse
  • looked at the MH mesh and measurement plugin once again...

2012-11-13

  • saw the stuff Klaus made
  • told Perttu I'm outta here
  • started writing Approach from the beginning
  • skimmed anatomy books meant for artists

2012-11-12

  • digged into anatomy
  • changed to biblatex

2012-11-06/09

  • wrote a lot of stuff in the Literature section
  • still unfinished :(
  • made a Kinfu scan of myself (with Tuukka's help)
  • processed this in MeshLab to get a mesh suitable for animation
  • used Pinocchio to auto-rig the mesh and animate it

2012-11-05

  • skimmed through Charpentier's dissertation and wrote about it
  • read about Pinocchio and compiled it

2012-11-02

  • at HiQ again
  • briefly chatted with Tassu

2012-11-01

  • writing...

2012-10-31

  • wrote more than usual
  • helped grade the graphics exams

2012-10-30

  • was at HiQ :)
  • re-read the Ahmed paper and wrote a general description
  • decided to include the Stoll paper and read it

2012-10-29

  • took the time to understand TSDF the best I can

2012-10-24

  • review my thesis with Tuukka
  • apply corrections

2012-10-23

  • read about history of ICP (Chen-Medioni, Besl-McKay)

2012-10-22

  • more literature (new articles + text in thesis)

2012-10-19

  • gave the first early version of my thesis to Tuukka

2012-10-17/18

  • was at the Range Imaging course
  • talked with Perttu about possible Ph.D. work
  • check the TC2 KX-16 and Bodymetrics products

2012-10-15

  • looked at finding the error between the point cloud and the mesh
  • edited base.obj to remove the skirt

2012-10-12

  • did the paperwork for getting approval for thesis topic

2012-10-10/11

2012-10-09

  • got the MH target prototype fully working

2012-10-08

  • started working on my own implementation of the MakeHuman target system in processing
  • made a small fix to OBJLoader

2012-10-03/05

  • wondered about MakeHuman vs. NITE licensing
  • made a python console plugin for MakeHuman
  • hacked with it, becoming a little desperate in the process

2012-10-02

  • totally froze my computer with Mitsuba: NVRM: GPU at 0000:01:00.0 has fallen off the bus.

2012-10-01

  • back at work
  • back to writing thesis
  • getting to know the internals of MH measurement plugin

2012-09-26/28

  • sick leave (wisdom teeth removal)

2012-09-21

  • read some of the MH mailing list backlog
  • got my hands dirty with Blender
  • started fleshing out thesis introduction

2012-09-20

  • attended Aalto Science Day

2012-09-19

  • got PCL compiled and started running my own build
  • contributed a bit to MakeHuman
  • helped test alpha7
  • fixed the censor UI

2012-09-18

  • found out Kinect doesn't work on USB3 (!)
  • found out OpenNI is problematic in many ways...
  • got my OpenNI installation working after I botched it yesterday (because I thought the drivers don't work while USB3 was the culprit)

2012-09-17

  • got stuck with most binaries I need having been compiled for glibc 2.14/2.15 while Debian has 2.13
  • compiled a lot of stuff by hand
  • failed to compile SimpleOpenNI and got frustrated
  • copied the whole Ubuntu installation and got most everything working (except Kinect)

2012-09-14

  • installed most software on Debian

2012-09-13

  • decided configuring Ubuntu once again majorly sucks
  • installed Debian testing
  • copied important stuff over
  • didn't write my thesis :(

2012-09-12

  • got in touch with Manuel Bastioni
  • joined the MakeHuman Google Group
  • got the new PC
  • installed Ubuntu 12.04

2012-09-11

  • researched MakeHuman

2012-09-10

  • decided to work on my thesis every day at about 13:30, requiring myself to have some improvement by 14:00
  • changed \fixme to use red color

2012-09-06/07

  • made changes to my thesis following the discussions on Wednesday
  • pretty stuck with MakeHuman
  • made plans about which lectures I'm gonna attend during the autumn

2012-09-05

  • had a meeting with Tassu about my thesis
  • investigated MakeHuman's pure python branch

2012-09-03/04

  • played with MakeHuman
  • tuned thesis

2012-08-30

  • more work on thesis
  • one more dataset run for Tommi
  • group meeting
  • made an appointment with Tassu for next Wednesday, 10am (about my thesis)

2012-08-29

  • worked on a my thesis (TOC, abstract)
  • still checked some minor details with Tommi's datasets
  • glanced on Texturing and Modeling

2012-08-28

  • read some papers
  • still worked on Tommi's stuff
  • decided to start working on my thesis' table of contents in the evening
  • played with MakeHuman

2012-08-27

  • worked on upscaling, video recording etc. related to Tommi's data
  • got the setup automated
  • made some runs for Tommi

2012-08-24

  • met with Tommi, hacked together

2012-08-23

  • implemented zmap2pcd
  • implemented camera matrix printing in hanfu

2012-08-22

  • got connected with Tommi and glanced through his papers

2012-08-21

  • Seppo said that Geforce cards can't be installed on Fujitsu machines like the one I have
  • got the EM-ICP/Softassign implementation on CUDA compiled
  • requires CUDA, liblapack-dev, libatlas-dev
  • If OpenGL version > 3.0 the point cloud viewer won't compile because of gluPerspective and gluLookAt. (these can be commented out, the viewer is unnecessary...)
  • wrote basic structure for the Approach chapter

2012-08-20

  • read some more of OpenNI and PCL source code
  • tried compiling the CUDA implementation of EM-ICP (problems with some libraries missing...)
  • Seppo tried to install a Geforce GTX 670 for me

2012-08-16

  • found out about different ICP versions
  • thought of re-orienting the point clouds using skeleton data in the limbsplit step

2012-08-15

  • finally got automated mesh extraction working
  • if kinfu resets just before the recording ends, nothing gets saved
  • it would be nice to get rid of some cruft (no need for visualization, etc)
  • decided to put the kinfu approach aside for now

2012-08-14

  • hacked hanfu
  • added the --save-mesh <file> cmdline parameter

2012-08-13

  • forked hanfu so I can hack my own stuff :)
  • added the --repeat commandline parameter
  • changed TSDF resolution so my measly 1GB of GPU memory will suffice for a while
  • looked at my first "limb" meshes generated by kinfu
  • found the head (it's obviously at index 0 whereas I started counting from 1 before)
  • realized that 1GB GPU memory is too little for kinfu, and got a GTX 670 ordered for me

2012-08-10

  1. frames (with body part segmentation) are grabbed into a timestamped directory in Processing (SkeletonPractice)
  2. body parts are separated into their own "animations" as limb01/*.pcd etc using the script limbsplit
  3. the animations are played like kinfu_app -pcd <path>/limb01 -pcd_fps <framerate>

2012-08-09

  • got kinfu_app to read .pcd directories I've generated
  • note: requires WIDTH 640\nHEIGHT 480 style header
  • loaned Texturing and Modeling: A Procedural Approach

2012-08-08

  • tried to read the mess that is kinfu source code, along with some other PCL components

2012-08-07

  • found and investigated Kintinuous
  • collected BibTeX references for articles read so far
  • had a discussion with Tuukka
  • decided to try running parallel kinfu instances for each limb (Idea: a single limb is close to a static object, which kinfu is good with. Since we already can segment the point cloud limb-wise, we can remove all other points. Moving the limb is pretty much equivalent to moving the camera, so this might work.)

2012-08-06

  • read some helpful material wrt. writing a Master's thesis
  • e.g. Timo Laakso: Miten kirjoitan diplomityön
  • created a basic LaTeX template for writing

2012-W27

  • worked with data in matlab
  • found out matlab isn't very useful
  • tried out matplotlib
  • read about NURBS
  • OpenNURBS is coming to PCL (exists in trunk)
  • played with ellipsoids
  • read the remaining unread papers
  • spent 1.5 days sick

2012-07-06 Friday

  • played with different algorithms in MeshLab
  • tried to get libfreenect-goodies working

2012-07-05 Thursday

  • tried to get libfreenect-goodies working

2012-07-04 Wednesday

  • implemented skeleton export to .pcd comments (now with joint orientations)

2012-07-03 Tuesday

  • read papers
  • started work on record/playback

2012-07-02 Monday

  • made this diary
  • recorded and toyed with .pcd files
  • learned to convert .pcd -> .obj
  • made my first mesh with meshlab

2012-W26

  • segmenting point cloud by bones (nearest bone)
  • drawing skeleton as cylinders fitted to the point cloud
  • played with color blending
  • colored cylinders with average cloud color
  • implemented .pcd saving

2012-W25

  • played around with SimpleOpenNI's User3D example
  • implemented cylinder drawing and started drawing bones with them
  • moved to room 2550

2012-W24

  • got PCL and kinfu working
  • other software installations
  • playing with OpenNI and Microsoft Kinect SDK
  • read some papers

2012-W23

  • got settled
  • read some papers (...)
Add a custom footer
Add a custom sidebar
Clone this wiki locally