Unexpected behavior with camera.draw_domain and camera.draw_grids
Hi yt users, I'm having a lot of trouble recreating a volume rendering I did a couple of weeks ago. Specifically, when I try to add a boundary box to my snapshot through draw_domain() or grids through draw_grids(), the transfer function rendering fades very significantly, regardless of alpha value, and it did not behave this way before. However, it does not fade for cam.draw_box(... , ... , np.array([0.1, 0.1, 0.1, 0.2]) . Since draw_domain is basically just a call to draw_box, I took a look in camera.py and did a bit of testing, and the only difference is that the color array is np.array([1.0,1.0,1.0,alpha]) (where alpha=0.2) for draw_domain, and that is what seemed to be causing so much fade in the case of draw_domain. I didn't experience this problem before. Also, I don't know much about how this color stuff works, but I was under the impression that only the alpha value determined how bright the box is compared to the rendering underneath. Furthermore, when I decrease the alpha value, the box fades, but so does the rendering underneath, which should not be the case (?). I see the same behavior when doing draw_grids. Here is the image I created a couple of weeks ago: http://i.imgur.com/R0uhz2D.png And here is the image that the same script (very nearly) creates now: http://i.imgur.com/xcTmCMC.png The rendering underneath the grids is nearly invisible in the second one. If I stack another draw_grids onto the image after this, for a different level range, the rendering goes completely invisible, and the first set of grids almost does-- again, even with alpha values of ~0.05. For reference, here is the script I'm using: ----------------------- from yt.mods import * pf = load("plt_derr_perr_grav_00020") field = "pressure" [variable definitions, cut out for clarity] pf.h pf.field_info[field].take_log=True tf = ColorTransferFunction((mi-1, ma+1), nbins=1.0e6) tf.sample_colormap(np.log10(2.0e21), 0.006, col_bounds=[mi,ma], alpha=[55.0], colormap="RdBu_r") tf.add_layers(6, mi=di, ma=da, col_bounds=[pu,da], alpha=26*na.ones(6,dtype='float64'), colormap="RdBu_r") [more layers, cut out for clarity] tf.grey_opacity=True c = [5.0e9, 5.0e9, 5.0e9] L = [0.15, 1.0, 0.4] W = (pf.domain_right_edge - pf.domain_left_edge)*1.25 Nvec = 768 cam = pf.h.camera(c, L, W, (Nvec,Nvec), transfer_function = tf, fields=[field], pf=pf, no_ghost=False) im = cam.snapshot(clip_ratio=6.0) im.add_background_color('black', inline=True) cam.draw_box(im, np.array([3.0e9, 4.0e9, 4.0e9]), np.array([7.0e9, 6.0e9, 6.0e9]), np.array([0.1, 0.1, 0.1, 0.2])) nim = cam.draw_grids(im, alpha=0.08, min_level=1) nim.write_png('v1.png') ----------------------- Am I doing something wrong or missing something obvious? Was the behavior of draw_grids and draw_domain changed with the new update? Thanks for your help! -Noel Scudder
Hi Noel, Thanks for bringing this up. I'm afraid I messed up (or really never had it right to begin with!) how rgba channels are blended when using grey_opacity=True renders. I've posted a PR here: https://bitbucket.org/yt_analysis/yt/pull-request/553/fix-for-grey_opacity-g... resolves this problem for me. I don't expect it to take too long to be reviewed and pulled in to the main repository, but if you are in a hurry feel free to pull down the changes and test it out yourself. If you aren't familiar on how to do that, please let me know. Best, Sam BTW, very nice volume rendering! On Tue, Jul 9, 2013 at 4:53 PM, Noel Scudder <noel.scudder@stonybrook.edu>wrote:
Hi yt users,
I'm having a lot of trouble recreating a volume rendering I did a couple of weeks ago. Specifically, when I try to add a boundary box to my snapshot through draw_domain() or grids through draw_grids(), the transfer function rendering fades very significantly, regardless of alpha value, and it did not behave this way before. However, it does not fade for cam.draw_box(... , ... , np.array([0.1, 0.1, 0.1, 0.2]) . Since draw_domain is basically just a call to draw_box, I took a look in camera.py and did a bit of testing, and the only difference is that the color array is np.array([1.0,1.0,1.0,alpha]) (where alpha=0.2) for draw_domain, and that is what seemed to be causing so much fade in the case of draw_domain. I didn't experience this problem before. Also, I don't know much about how this color stuff works, but I was under the impression that only the alpha value determined how bright the box is compared to the rendering underneath. Furthermore, when I decrease the alpha value, the box fades, but so does the rendering underneath, which should not be the case (?).
I see the same behavior when doing draw_grids. Here is the image I created a couple of weeks ago: http://i.imgur.com/R0uhz2D.png And here is the image that the same script (very nearly) creates now: http://i.imgur.com/xcTmCMC.png The rendering underneath the grids is nearly invisible in the second one. If I stack another draw_grids onto the image after this, for a different level range, the rendering goes completely invisible, and the first set of grids almost does-- again, even with alpha values of ~0.05.
For reference, here is the script I'm using:
----------------------- from yt.mods import * pf = load("plt_derr_perr_grav_00020") field = "pressure" [variable definitions, cut out for clarity]
pf.h pf.field_info[field].take_log=True
tf = ColorTransferFunction((mi-1, ma+1), nbins=1.0e6)
tf.sample_colormap(np.log10(2.0e21), 0.006, col_bounds=[mi,ma], alpha=[55.0], colormap="RdBu_r")
tf.add_layers(6, mi=di, ma=da, col_bounds=[pu,da], alpha=26*na.ones(6,dtype='float64'), colormap="RdBu_r")
[more layers, cut out for clarity]
tf.grey_opacity=True
c = [5.0e9, 5.0e9, 5.0e9] L = [0.15, 1.0, 0.4] W = (pf.domain_right_edge - pf.domain_left_edge)*1.25 Nvec = 768
cam = pf.h.camera(c, L, W, (Nvec,Nvec), transfer_function = tf, fields=[field], pf=pf, no_ghost=False) im = cam.snapshot(clip_ratio=6.0)
im.add_background_color('black', inline=True) cam.draw_box(im, np.array([3.0e9, 4.0e9, 4.0e9]), np.array([7.0e9, 6.0e9, 6.0e9]), np.array([0.1, 0.1, 0.1, 0.2]))
nim = cam.draw_grids(im, alpha=0.08, min_level=1)
nim.write_png('v1.png') -----------------------
Am I doing something wrong or missing something obvious? Was the behavior of draw_grids and draw_domain changed with the new update? Thanks for your help!
-Noel Scudder
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
Hi Sam, Great, this seemed to have worked. With a bit of tweaking, I'm able to basically recreate my original rendering: http://i.imgur.com/D8z0oFg.png Thanks for fixing this! -Noel On Wed, Jul 10, 2013 at 12:54 PM, Sam Skillman <samskillman@gmail.com>wrote:
Hi Noel,
Thanks for bringing this up. I'm afraid I messed up (or really never had it right to begin with!) how rgba channels are blended when using grey_opacity=True renders. I've posted a PR here: https://bitbucket.org/yt_analysis/yt/pull-request/553/fix-for-grey_opacity-g... resolves this problem for me. I don't expect it to take too long to be reviewed and pulled in to the main repository, but if you are in a hurry feel free to pull down the changes and test it out yourself. If you aren't familiar on how to do that, please let me know.
Best, Sam
BTW, very nice volume rendering!
On Tue, Jul 9, 2013 at 4:53 PM, Noel Scudder <noel.scudder@stonybrook.edu>wrote:
Hi yt users,
I'm having a lot of trouble recreating a volume rendering I did a couple of weeks ago. Specifically, when I try to add a boundary box to my snapshot through draw_domain() or grids through draw_grids(), the transfer function rendering fades very significantly, regardless of alpha value, and it did not behave this way before. However, it does not fade for cam.draw_box(... , ... , np.array([0.1, 0.1, 0.1, 0.2]) . Since draw_domain is basically just a call to draw_box, I took a look in camera.py and did a bit of testing, and the only difference is that the color array is np.array([1.0,1.0,1.0,alpha]) (where alpha=0.2) for draw_domain, and that is what seemed to be causing so much fade in the case of draw_domain. I didn't experience this problem before. Also, I don't know much about how this color stuff works, but I was under the impression that only the alpha value determined how bright the box is compared to the rendering underneath. Furthermore, when I decrease the alpha value, the box fades, but so does the rendering underneath, which should not be the case (?).
I see the same behavior when doing draw_grids. Here is the image I created a couple of weeks ago: http://i.imgur.com/R0uhz2D.png And here is the image that the same script (very nearly) creates now: http://i.imgur.com/xcTmCMC.png The rendering underneath the grids is nearly invisible in the second one. If I stack another draw_grids onto the image after this, for a different level range, the rendering goes completely invisible, and the first set of grids almost does-- again, even with alpha values of ~0.05.
For reference, here is the script I'm using:
----------------------- from yt.mods import * pf = load("plt_derr_perr_grav_00020") field = "pressure" [variable definitions, cut out for clarity]
pf.h pf.field_info[field].take_log=True
tf = ColorTransferFunction((mi-1, ma+1), nbins=1.0e6)
tf.sample_colormap(np.log10(2.0e21), 0.006, col_bounds=[mi,ma], alpha=[55.0], colormap="RdBu_r")
tf.add_layers(6, mi=di, ma=da, col_bounds=[pu,da], alpha=26*na.ones(6,dtype='float64'), colormap="RdBu_r")
[more layers, cut out for clarity]
tf.grey_opacity=True
c = [5.0e9, 5.0e9, 5.0e9] L = [0.15, 1.0, 0.4] W = (pf.domain_right_edge - pf.domain_left_edge)*1.25 Nvec = 768
cam = pf.h.camera(c, L, W, (Nvec,Nvec), transfer_function = tf, fields=[field], pf=pf, no_ghost=False) im = cam.snapshot(clip_ratio=6.0)
im.add_background_color('black', inline=True) cam.draw_box(im, np.array([3.0e9, 4.0e9, 4.0e9]), np.array([7.0e9, 6.0e9, 6.0e9]), np.array([0.1, 0.1, 0.1, 0.2]))
nim = cam.draw_grids(im, alpha=0.08, min_level=1)
nim.write_png('v1.png') -----------------------
Am I doing something wrong or missing something obvious? Was the behavior of draw_grids and draw_domain changed with the new update? Thanks for your help!
-Noel Scudder
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
participants (2)
-
Noel Scudder -
Sam Skillman