20

Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

  • Upload
    others

  • View
    6

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave
Page 2: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

2 2  |  P a g e  

 

Getting  started  

One  of  the  most  fundamental  factors  for  getting  a  good  3D  model  reconstruction  from  photos  is  to  take  GOOD  photos  and  ENOUGH  photos  that  have  GOOD  overlap.  Please  refer  to  the  end  of  the  document  for  tips  on  good  shooting  strategies.  

Getting  started  -­‐  A360  Sign  in  

To  get  access  to  the  Photos-­‐to3D  Autodesk®  ReCap™  Photo  online  service,  you  will  need  an  account  on  Autodesk’s  online  cloud  platform  A360        https://360.autodesk.com    

To  run  A360,  you  must  use  Google  Chrome  as  an  internet  browser.  Safari  and  Internet  Explorer  are  not  supported.  Once  you  have  A360  running  in  your  browser,  do  the  following:  

• Login  on:  https://360.autodesk.com/landing  with  your  A360  account  • If  you  do  not  have  one,  hit  the  Create  button.  

   

 

   

Page 3: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

3 3  |  P a g e  

 

Basic  Workflow:  From  photos  to  a  3D  model    

Upload  photos  and  submit  to  invoke  the  photo-­‐to-­‐3D  service  

1. In  A360,  select  Documents  (top  left  of  your  screen),  and  create  a  new  folder  for  your  project.    2. Upload  the  photos  in  that  folder.  Note  that  the  service  currently  can  only  process  JPEG  images.  

It   is   currently  not  possible   to   create  a   folder  on   the   fly,  while  uploading   the  photos.  You  will  need   to  make   one   prior   to   the   upload   process.   It   is   a   best   practice   to   always   create   a   new  folder  for  each  project.  This  will  make  your  photos  manageable.  Otherwise  everything  will  be  landing   in   the  main  directory  of   the   folder  and  soon  you  will  have  a  mess  of  photos  and   .rcp  files  from  various  projects.    

               Select  all  uploaded  photos  and  right  click  over  them  to  get  the  Create  3D  model  option.    To  select  all  photos  either  check   the   first  one  and   then   the   last  one  holding   the  Shift  key,  or  

check  the  little  folder  icon  on  top  left  of  the  Documents  screen.      Your   selection   must   only   contain   photos,   otherwise   the   ‘Create   3D   model’   option   will   not  show!    

3. After  all  your  photos  have  been  selected,   right  click  on  them  and  click   the   ‘Create  3D  model’  button.    

 

 

Page 4: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

4 4  |  P a g e  

 

4. A  Photo-­‐to-­‐3D  pop-­‐up  window  will  open.  On  the  left  side  of  the  window,  you  will  find  various  settings  specific  to  the  3D  reconstruction.    

   Here  you  should  do  the  following:  

• Name   your   project   –     At   the   top   left   in   the   side   bar   there   is   a   place   to   name   your  project.  

• Select  a  mesh  density  /quality    There  are  3  choices  for  mesh  densities/qualities  for  the  generated  3D  models:  

§ Draft   –   you   can   use   this   mode   to   get   a   quick   result   and   see   if   your  stitching   (3D  model  creation)  will  be  successful.     In  draft  mode,  you  can  view  only  the  reconstruction,  but  you  cannot  export  the  mesh.  (The  export  options  are  greyed  out  in  this  mode)  § Standard  or  Maximum  –  These  mesh  densities  give  you  a  high  quality  mesh  model  that  can  be  exported  in  various  formats.      

• Select  the  file  formats  in  which  you  wish  to  save  the  resulting  3D  model.  You  can  save  as  .OBJ,  .FBX,  .IPM  and  .RCS  (Autodesk  point  cloud  format),    format.  A  .RCP  file  will  be  generated  automatically  with  every  3D  model  reconstruction.  This  is  the  native  Reality  Capture  Project  file  and  it  is  that  file  that  you  will  be  opening  to  view  in  your  browser.  

When  done  with   your   selections,   submit   your  project   by   clicking  on   the   rocket   icon   (top  right  or  bottom  right  part  of  the  screen).  Once  you  have  clicked  on   ‘Submit  Project,’  wait  until  this  message  appears:  

 This  means  that  your  data  is  sent  to  the  cloud  for  calculation,  and  that  you  will  receive  an  email  that  will  inform  you  when  the  calculation  is  finished.  Note  that  the  email  will  be  sent  to  the  email  address  with  which  you  logged  in  /  registered  on  A360.      You  will   notice   that   upon   the   submission,   a   new   .RCP   icon   is   added   to   your   folder   as   a  placeholder,  indicating  work  in  progress.  

Page 5: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

5 5  |  P a g e  

 

   From   this   moment   on,   the   cloud   service   is   calculating   the   reconstruction   of   the   3D   model.  When  the  service  is  done  calculating,  you  will  receive  an  email  informing  you  of  the  success  of  the  reconstruction  and  a  link  to  the  3D  model.  This  will  take  some  time  (especially  if  you  selected  Standard  or  Maximum  mode)  so,  there  is  no  need  to  wait  here.    You  can   leave  your  A360  space  and  continue  working  on  something  else,  until  you  receive  an  email  from  the  service.    The  email  will  be  sent  by  ‘[email protected]’.    Note:  if  you  selected  ‘Draft’  mode,  the  calculation  process  will  be  fairly  quick.  However,  if  you  selected  ‘Standard’  or  “Maximum’  mesh  resolution,  the  process  can  take  from  30  minutes  up  to  a  couple  of  hours,  depending  on  the  number  of  photos  you  submitted  and  their  resolution.  For  high  quality  mesh  generation,  you  will  probably  be  dealing  with  70+  photos,  so  expect  that  this  process  takes  at  least  an  hour.    Also  note  that  the  service  is  currently  optimized  for  projects  with  less  than  250  photos.  

Preview  your  3D  model  results  in  A360  web  viewer  and  download  it  on  your  desktop  

In  the  email  there  will  be  a  hyperlink  that  when  clicked,  opens  the  .RCP  model  directly   in  the  A360  web  viewer.  Alternatively,  you  can  open  the  model   from  within  A360   folder.  When  the  .RCP  file  is  ready,  you  will  see  a  thumbnail  image  of  your  model  there.    

                                                                                                                                 

5. If  you  are  working  from  within  your  A360  folder,  click  on  the  .RCP  icon  and  this  will  initiate  the    Web  3D  Viewer  where  your  scene  will  be  displayed.  

Page 6: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

6 6  |  P a g e  

 

                                                                         

A  few  things  to  note  here:  You  might  get  confused  when  you  first  look  at  the  generated  scene,  as  it  probably  will  not  look  like  the  object  you  tried  to  reconstruct  (a  warrior  statue  in  the  example  case).  This  is  because  the   software   calculates   the  entire   scene  around   the  object   as  well,   not   just   the  object   itself.  You  will  need  to  zoom  inside  the  scene  and  find  the  target  object.  You  will  need  to  clean  up  the  rest  of  the  scene  that  you  don’t  desire  in  any  mesh  software  (3dsMax,  Maya,  Mudbox  etc.)  

 Secondly,  as  our  service  is  young  and  we  want  to  track  its  accuracy,  You  will  be  prompted  to  let  us  know  if  you  are  satisfied  with  the  reconstruction.      To  navigate  the  3D  scene,  you  can  use  either  the  navigation  tools  displayed  in  the  side  bar  on  the  right,  or  the  mouse:  left  button  for  orbiting,  middle  button  for  zoom  and  middle  mouse  +  hold  for  pan.    

Page 7: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

7 7  |  P a g e  

 

You  can  view  the  generated  3D  model  as  a  textured  model  (Default  setting)  or  as  mesh  only.  To  switch  between  the  two,  click  on  the  monitor  icon  in  the  menu  on  the  right  side  and  select  the  globe  icon  for  mesh  only  display.  To  come  back  to  texture  display,  click  again  on  the  globe  icon.      

                                                           

 

Figure  1  -­‐  Switch  to  mesh  mode  

You  can  also  display  the  camera  positions  where  the  photos  were  taken,  by  clicking  the  camera  icon  in  this  sub-­‐menu.    

Resubmitting  your  scene:    

You  can,  at  any  time,  resubmit  your  scene  to  be  calculated  again.  There  are  various  cases  in  which  you  will  need  to  resubmit  your  scene:  

a) If  you  first  started  with  ‘Draft’  mode,  and  after  checking  the  result  in  the  Web  3D  Viewer,  you  want  to  create  a  higher  mesh  quality  (Standard  or  Maximum  Resolution).      

b) After  you  have  manually  stitched  photos  that  have  not  been  included  automatically.  These  would  be  then  considered  in  a  new  reconstruction.  

To  resubmit  a  scene,  go  to  the  ‘Action’  menu  while  in  the  Web  3D  Viewer.  You  can  also  access  this  by  right-­‐clicking  the  .RCP  icon.      

Page 8: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

8 8  |  P a g e  

 

                                                             

When   you   click   ‘Resubmit   3D   model’,   the   ReCap   photo   settings   UI   opens   again   and   here   you   can  change  the  settings  to  Standard  or  Maximum  quality,  and  pick  which  file  formats  you  wish  to  export  to.  

Then,  you  click  the  Submit  button  again  to  restart  a  new  calculation.  

After  resubmitting  your  scene  in  Standard  /  Maximum  resolution,  you  will  again  receive  an  email.  The  model  you  will  see  on  the  screen  will  be  visibly  better  quality,  although  you  should  note  that  you  are  still   only   viewing  a  proxy  of   the  actual  mesh,  not   the   full   resolution  of   the  mesh,  even   if   the  mesh  density  you  set  was  Standard  or  Maximum    

The  following  message  pops  up  when  opening  the  Web  3D  viewer:  

 

 

Its  important  to  understand  that  this  is  just  a  viewing  limitation  that  does  not  give  justice  to  the  quality  of   the   3D  meshes   that   are   created.   The   generated   .OBJ   or   .FBX  meshes   are   high  quality  meshes.   To  view  those  high  resolution  meshes,  download  the  .OBJ  or  .FBX    and  use  applications  such  as  3ds  Max,  Mudbox,  Maya  or  similar  to  view  or  further  edit  the  meshes.  

 

Page 9: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

9 9  |  P a g e  

 

MANUAL  STITCHING  

The  above-­‐described  workflow  is  straightforward  process  and  is  all  you  need  to  do  if  you  had  good  number  and  quality  of  photos.  However,  until  you   learn  how  to  take  photos  specific   for  photogrammetry  and  gather  experience  based  on  what  works  best,   you  will  probably  need   to  do  some  manual  work  before  arriving   to  a  good  quality  mesh  output.    When   there   are   issues  with   your   photos,   after   the   scene   is   reconstructed   for   the   first   time,   you  might  see  that  you  have  a  few  ‘unstitched’  photos,  displayed  in  the  side  bar  (usually  you  need  to  scroll   down).   Unstitched   photos   are   photos   that   the   service   couldn’t   take   into   account   for   the  model  generation,  because  they  have  little  similarity  with  the  rest  of  the  photos  (for  example,  the  contrast   between   two  photos   is   too   high,   or   the   overlap   is   too   small   etc.).   Adding   those   photos  might  increase  the  quality  of  the  generated  3D  model,  but  it  requires  a  manual  stitching  process.    

Manual  stich:  

Manual   stitching  means   that  you  will  manually  define  matching  of  points  between  an  unstitched  photo  and  stitched  photo  so  that  the  system  takes  the  unstitched  photo  in  account.  

To  start  a  manual  stitch,  from  the  left  side  bar,  scroll  down  to  see  unstitched  images,  pick  one  and  select  STITCH  on  the  top  right  of  the  screen  

 

 

In  the  new  manual  stitch  window,  the  photo  that  you  selected  to  be  stitched  will  be  displayed  on  the  left  side  on  the  screen.  On  the  right  side,  you  will  be  offered  a  choice  of  multiple,  already  stitched  photos  to  pick  from.  You  should  pick  an   image  that  has  many  similar  points  with  the  unstitched  one.  

Page 10: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

10 10  |  P a g e  

 

 

Expand  the  little  icon  on  the  left  corner  of  the  right  screen  to  see  the  photos  to  choose  from.  

 

You  should  now  start  identifying  matching  points  between  the  two  photos  that  are  common  in  order   to   help   the   system   understand   where   the   unstitched   photo   fits.   Defining   4   matching  points  in  total  is  usually  good  practice.  

When   you   click   on   a   point   you  wish   to   set   as   common  between   the   two  photos,   the   cursor  turns   into  a  magnifying  glass   that  helps  more  precisely  define   the  point.   You  can  change   the  magnification   factor  with   your  mouse  wheel.   Start   finding   a   point   on   the   left   click.   It  will   be  marked  as  Point  1.  Then  try  to  find  the  same  point  on  the  photo  on  the  right  side.  That  will  also  be   marked   as   Point   1.   Continue   until   you   do   this   with   4   matching   points.   You   should   see  something  like  this  when  you  are  done.  

                                                   

Page 11: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

11 11  |  P a g e  

 

On   the   bottom   left   of   the   screen   you  will   also   have   a   display   of   all  matching   points   and   the  corresponding  selections.    To  view  all  points  you  have  identified,  hover  over  the  first  icon.  Then  you  can  click  on  any  of  the  other  points  to  see  the  corresponding  matching  points.  

                                                         

 

When  done,  click  on  the  Actions  menu,  select  Resubmit  model  and  in  the  settings  window  click  Submit  project.  The  service  will   restart   the  calculation  and  you  will   receive  an  email  when   its  done.  

ADVANCED  FEATURES  AND  WORKFLOWS  

ReCap  Photo  has  advanced  features  and  those  are  separated  from  the  straightforward  photo-­‐to-­‐3D  process.   You   can   see   the   additional   advanced   tools   by   clicking  on   ‘advanced   tool’   button  on   the  bottom  left  of  the  sidebar  in  the  initial  settings  screen.    

     

This   will   invoke   new   tools   that   will   appear   in   the   blue   ribbon   on   the   top   of   the   screen:  Registration  and  Texture.  These  tools  should  be  used  prior  to  a  3D  reconstruction,  at  the  very  beginning  of  the  process.  

 

Registration  tools  

In   the   registration   panel   you   can   do   two   things:   work   with   survey   points,   and   use   reference  distances  to  scale  your  model.  

Survey  Points  

Survey  points  are  collected  on-­‐site  using  ‘Total  Stations’  or  laser  scanners.  They  are  ‘hard’  3D  points  used  in  the  stitching  process  to  make  it  more  robust,  and  to  scale  or  locate  your  3D  model.  To  add  a  survey  point,  select  a  point  on  the  photo  for  which  you  know  the  coordinates,  and  once  clicked,  fill  in  the  X,Y  and  Z  coordinates  on  the  bottom  left  of  the  screen:  

Page 12: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

12 12  |  P a g e  

 

                     

 

Reference  distances  

Photogrammetry   does   not   have   a   sense   of   scale.  When  we   create   a   3D  model   from   photos   the  actual  size  of  the  reconstructed  3D  object  is  unknown.  If  you  know  the  distance  between  points  on  your  original  object  you  can  set  a  distance  to  your  model  and  thus  scale  the  model  to  its  actual  size.  You  can  do  this  with  the  help  of  the  reference  distance  tool  which  is   located  on  the  bottom-­‐right  corner  of  the  Registration  screen.    

 

Note  that  you  should  do  this  PRIOR  to  submitting  your  photos  for  3D  reconstruction.  

To  start  the  process  for  setting  a  reference  distance,  select  the  advanced  tools  in  the  side  bar  and  then,  in  the  blue  ribbon,  click  on  Registration.  

 

A  UI  with  two  windows,  similar  to  the  manual  stitch  window,  will  open.  Here  you  pick  one  image  in  the   left   pane   that   shows   the  part   of   the  object   for  which   you   know   the   reference  distance,   and  then  select  another  image  in  the  right  pane  where  you  can  also  see  the  same  reference  for  which  you  have  a  distance  value.  

Page 13: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

13 13  |  P a g e  

 

 

The   process   will   require   you   to   identify   a   pair   points   in   both   images   and   then   set   a   reference  distance  between  them.  

Start  by  clicking  in  the  left  pane  to  create  Point  1.  In  the  same  pane  create  point  2.    

 

Now  move  to  the  right  side,  and  in  the  right  pane  match  point  1  and  point  2.   It   is  very  important  that  you  activate  point  1  in  the  left  pane  before  selecting  the  corresponding  point  in  the  right  pane.  To  do  that  just  move  over  point  1  in  the  left  pane  and  click  on  it.  It  will  become  active.    

Page 14: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

14 14  |  P a g e  

 

 

Do  the  same  for  point  2.  Click  first  in  the  left  pane  on  Point2  to  activate  it  and  then  click  in  he  right  pane  on  the  same  point.  

You  should  end  up  with  point  1  and  2  in  both  the  left  and  the  right  pane.  

 

Only  now  should  you  click  on  the  reference  distance  tool,   located  in  the  lower  right  corner  of  the  screen.  When  opened,  click  on  the  tape  measure  icon  and  then  click  once  on  Point  1  in  one  of  the  panes,  and  then  on  Point2.  The  distance  field  in  the  measure  UI  will  now  be  active,  and  you  should  type  the  known  distance  value  in  there.  To  finish,  select  the  TAB  key  (that  is  Enter  for  browsers).  

 

What  you  should  end  up  with  is  a  blue  line  that  connects  point  1  and  point  2  in  both  of  the  panes.  

Page 15: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

15 15  |  P a g e  

 

                             

Note  that  in  the  current  release  we  do  not  let  you  define  the  units  for  the  distance.    

Texture  

By   default,   ReCap   photo   uses   all   the   photos   both   for   the   geometry   creation,   as   well   as   for   the  texture  creation.  If  you  are  an  advanced  user,  understand  photogrammetry  well  and  have  decided  to  do  some  work  in  photoshop  to  make  the  texture  look  prettier  for  visualization  purposes,  you  will  be  able  to  decide  to  keep  the  original  images  for  the  geometry  creation  only  (these  are  displayed  as  Calibration   images)  and  use  the  new,  visually   improved  texture   images  for  calculating  the  texture  (these  are  displayed  as  Texture  images).    

 

In  the  Calibration  images  row  you  see  all  original  images  used  for  mesh  creation.    

In  the  Texture  images  row  you  see  those  used  for  texture  creation.    

They  are  initially  the  same,  but  the  texture  images  can  be  either  turned  off  by  simple  click  over  them  (this  is  when  you  don’t  want  them  to  participate  in  the  texture  creation  because  their  colors  are  off)  OR  you  can  replace  them  all  with  new  ones,  after  treating  them  in  a  photo  editing  software,  by  clicking  the  ‘use  different   texture   images’  option  on   the   left  of   the  pane.   Its   very   important   to   leave   the  original  name  of  the  photo  in  the  edited  photo  and  just  add  an  index  to  it  (a,x,etc.)  

When  replacing  a  texture  image  with  a  new  image,  the  left  pane  will  show  the  photo  that  was  used  for  mesh  creation,  while  the  right  pane  the  one  that  will  be  used  for  texture  creation  

Page 16: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

16 16  |  P a g e  

 

 

 

Known  issues,  limitations  and  bugs  

 

Limitations  

§ BROWSER:  You  will  need  to  use  Google  Chrome  as  the  browser.  We  are  using  a  Web  GL  based  3D  model  viewer  to  preview  3D  models  in  A360  and  thus  Safari  and  Internet  Explorer  are  not  supported.  

§ Try  not  to  exceed  200  photos  per  reconstruction.  § There  might  be  a  5  min  time  lag  between  the  moment  you  receive  the  e-­‐mail  notifying  you  that  

the  3D  model   is   ready,  and  the  moment  your  project   files  are  actually  available   in  your  A360  folder.    

§ GoPro  cameras  are  not  yet  supported  by  ReCap  Photo.  They  use  a  fish-­‐eye  lens,  which  requires  a  different  mathematical  model  than  the  one  currently  implemented  in  ReCap  Photo.  

Bugs  

§ If   you   are   using   Mac,   you   might   run   into   an   issue   that   the   .OBJ   and   .FBX   after   being  uncompressed  are  not  readable   in  Maya  or  3dsMax.  We  found  out  that  some  decompression  software  turns  them  into  ‘No  access’  files.  Go  to  the  properties  of  the  unzipped  files  (Get  info  on  rmc)  and  change  the  permissions  from  ‘No  access’  to  ‘Read  and  write’.    

§ When  creating  a  new  folder  on  A360,  sometimes  the  folder  does  not  show  up  immediately.  Try  to  refresh  the  page  few  times,  until  it  appears.    

3D  Viewer  

As  previously  mentioned,  note  that  the  viewer  we  use  on  A360  to  preview  the  generated  3D  scene  has  limitations,  and  thus  what  you  will  be  viewing  on  screen  is  only  a  proxy  and  will  not  do  justice  to  the  actual  quality  of  the  mesh  and  texture,  when  processed  in  Standard  or  Maximum  Resolution.  You  can  see  the  high  quality  results  once  you  import  the  exported  meshes  in  Maya,  3DsMax  etc.  

 

Enjoy  using  ReCap  Photo  and  please  send  any  questions,  suggestions  and  wishes  on  the  ReCap  Photo  forum      http://feedback.autodesk.com/cloudservices/products/cloudservices_recap  

 

Your  ReCap  Team  

 

ADDENDUM  

How  to  take  good  photos  for  photogrammetry  

Page 17: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

17 17  |  P a g e  

 

What  makes  photos  good  for  photogrammetry?  

The   photogrammetry   process   looks   at   series   of   photos   and   reconstructs   the   camera   location,  orientation,  lens  type  and  lens  distortion. It  does  that  by  comparing  and  matching  pixels/points  across  photos  in  the  Entire  Scene  and  match/  triangulate  on  visual  features.  This  means  that  you  should  follow  the  following  guidelines:  

ENVIRONMENT  (Lighting  conditions)  

You  should  avoid  any  environment  in  which  the  light  creates  strong  contrast  shadows  over  the  object  you  shoot.  Shadows  are  enemy  to  the  reconstruction  process.  

• If  you  shoot  in  interior  o The  object  must  be  placed  in  a  room  with  diffuse  light  that  does  not  produce  contrast  light  

or  shadow.    o Do  not  use  the  flash.  

• If  you  shoot  outside  o Avoid  shooting  at  midday  under  strong  sun.  Best  times  are  early  in  the  morning  or  later  in  

the  afternoon  

EQUIPEMENT  

• You   can   get   good   results   with   any   camera.   However,   plastic   lenses   only   go   so   far   in   terms   of  sharpness.  If  you  are  serious  about  this  process  a  good  lens  makes  a  difference.  

• Try   to   shoot   the  photos  with  a   fixed   lens.   Fixed   lens  have  no   focal   change  which   is   good   for   the  photogrammetry  process.  A  50ml  lens  is  good.  

• Whenever  possible,  shoot  with  a  tripod  and  if  possible  a  clicker.  You  need  this  because  in  order  to  make   the   entire   photo   sharp,   you  will   have   to   set   a  wide  depth  of   field.   This  means   that   a   very  small  amount  of  light  will  be  coming  in  the  camera  so  you  will  need  to  set  your  shutter  speed  long.  This  makes  even  a  tiniest  of  shake  of  the  camera  result  blurry  photos.  The  tripod  and  a  clicker  will  assure  sharp  photos.    

SHOOTING  STRATEGIES  –  SCENE  SET  UP  

• Make  sure  that  he  object  you  shoot  is  always  in  the  middle  of  the  photo  and  fills  most  of  the  frame    • Use  rich  background.  If  you  shoot  a  single  object,  avoid  placing  them  on  a  monochromatic  surface.  

If   the   surface   is  monochromatic,   add  many   other   smaller   objects   around   the   target   object.   The  algorithms  need  as  many  different  references  in  the  space  to  do  better  calculation.  

• Avoid  green  photographers  screens  (ChromaKey).  Photogrammetry  doesn’t  do  well  with  them.  • Don’t  move  objects  around  during   the   shoot  and  don’t  have  people  walk  around.  Movements   in  

the  photos  create  uncertainty  in  the  reconstruction.  • Shoot  all  around  the  object  to  cover  each  aspect  of   it  with  multiple  overlapping  photos.  Shooting  

each  5-­‐15  degrees  on  the  same  height  of  the  object  is  a  good  practice.  If  you  want  to  have  a  good  level   of   detail,   you   should   shoot   at   2-­‐3   different   heights   of   the   object,   ending   up   with   70-­‐100  photos  for  a  small  to  medium  object  and  180-­‐200  photos  for  a  building.  

Page 18: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

18 18  |  P a g e  

 

                     • Take  enough  photos.  It  is  not  enough  to  have  many  photos,  but  they  must  also  have  a  good  overlap  

between  them.   If   the  overlap   is   too  small  or  non-­‐existent,   the  software  might  not   take  an  entire  cluster   of   photos   in   account   for   the   reconstruction,   and   the   reconstruction  will   be   inferior   to   its  potential.  

 • When   shooting   Interiors,   follow   the   directions   on   the   diagram  below.   Instead   of   standing   in   the  

middle  of  the  room  and  rotating  around  your  own  axis  to  shoot  the  photos,  move  along  the  walls  to  shoot  in  the  other  direction  

•  

When  shooting  Exteriors,  again,  avoid  standing  in  one  spot  and  shooting  ‘panorama  like’  spherical  photos  but  move  along  the  façade  with  the  camera.  

 • Don’t   have   people   walking   cross   the   scene   while   the   photos   are   taken.   Anything   that   makes   a  

movement  in  the  scene  makes  it  hard  for  the  algorithm  to  compare  pixels  between  the  photos    

Page 19: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

19 19  |  P a g e  

 

PHOTO  QUALITY  

• The  photos  must  be  in  focus  SHARP  and  CRISP  throughout  the  entire  scene  (  not  just  the  object  of  desire)   as   we   compare   pixels   from   the   entire   photo.   If   you   happen   to   have   few   blurry   photos  among  the  entire  set,  remove  them  before  making  the  reconstruction.  

• Avoid   blurriness   in   any   part   of   the   photos   (ah,   that   beloved  Depth   of   field   effect).   That  will   not  work  very  well  for  photogrammetry.  Do  not  assume  that  all  is  well  if  the  object  you  try  to  shoot  is  sharp.  We  calculate  the  entire  scene,  not  just  the  target  object  and  any  blurriness  will  confuse  the  algorithm.  

• Do  not  change  the  zoom  if  you  don’t  have  a  fixed   lens.  Rather  move  towards  and  away  from  the  object.  

• Do  not  change  exposure  during  the  same  acquisition.    • Set   ISO  to  100,   it  works  on  pretty  much  any  reconstruction.  Or  simply  keep  the   ISO   low.  High   iso  

brings  noise  which  should  be  avoided.  • The  ReCap  Photo  service  currently  accepts   JPEG   format  only.  However,   if  you  shoot   in  RAW,  you  

will  have  the  ability  to  do  some  changes  in  the  RAW  images  that  WILL  affect  the  quality  of  the  3D  model.  You  can  always  re-­‐export  JPEG’s  from  the  Raw  images  using  photo  editing  sowftare  such  as  Adobe  Photoshop  or  Bridge.  

 

KNOWN  TECHNOLOGICAL  LIMITATIONS  OF  PHOTOGRAMMETRY  

 The  photogrammetry  process  does  not  work  on  shiny,  transparent  or  highly  glossy  objects.  This  is  due  to   the   fact   that   these   reflect   the   surrounding  which  make   the   same   point   on   one   photo   to   be   in   a  different  color  on  another  photo,  making  it  impossible  for  the  algorithms  to  calculate  the  scene.    

 

 

 

 

 

 

 

 

 

 

 

Page 20: Getting)started) - Jarod Schultz · 2013-05-23 · 2 2"|Page" " Getting)started) One"of"the"most"fundamentalfactorsforgettingagood3 Dmodel"reconstruction"from"photos"is"totake" GOODphotos"and"ENOUGHphotos"thathave

20 20  |  P a g e  

 

 

 

For   any   comments   or   suggested   changes   or   additions   to   this   document,   feel   free   to   drop   a   note   to  [email protected]