Jump to content
Hash, Inc. Forums
R Reynolds

Netrender first frame strangeness

Recommended Posts

Before I begin trying to find the source of this bug I'd like some opinions as to what might be happening.

Attached are two "versions" of four sequential TIF files (545 through 548) from an on-going Netrender job. If there is a "b" (for bad) in the filename it is the first frame rendered by any one of the my four cores. If there is a "g" (for good) it was either not a Netrender first frame or it could be any frame rendered within A:M itself. I've numbered the files so you can sequentially show them bad, good, bad, etc. When I cycle through them on my monitor, the bad ones seem a bit darker and noisier but nothing that says "obviously this is what's wrong"; they're just poorer quality. All subsequent Netrender images are totally fine. The work-around is trivial but wastes some render time.

Debug will take stripping down the choreography until this stops happening but I'm open to suggestions as what to change/remove first.

 

01_b0545.tif 02_g0545.tif 03_b0546.tif 04_g0546.tif 05_b0547.tif 06_g0547.tif 07_b0548.tif 08_g0548.tif

Share this post


Link to post
Share on other sites

I have had a first frame problem previously...

https://reports.hash.com/view.php?id=6914

 

However, the two frames I downloaded for comparison from your post appear so dark that I can't see a difference between them.

Are they really supposed to be that dark?

This brings me to another issue: Are we sure your is monitor set right?

Here is a photo of my monitor displaying two of your frames along side a standard test image I downloaded.

I feel like my monitor is within normal bounds because the test image appears generally appropriate and i can distinguish all but the darkest two shades of gray in the test bars.

MonitorShot.jpg

external-content.duckduckgo.com.jpeg

Share this post


Link to post
Share on other sites

 

Here is a histogram of one of the frames. Is it really intended to just use the darkest half of the range?

image.png

Share this post


Link to post
Share on other sites

I remember going down this road with you before when I submitted my contest image. I was able to reproduce your monitor photo results almost exactly with my image also displayed beside a calibration image.

But it's not like I'm going out of my way to produce dark images, it just depends on the subject matter because my lighting setup never changes. The render shows the unlit interior of a diner on a sunny day. The first attached image shows the driver of a cab (partially reflected in the diner's window) standing outside using the same sunny day lighting. Next to him is my reference white sphere. The RGB value on the sphere's surface facing the sun is 255, 255, 255 and the surrounding area somewhat saturated as I suspect it would be to your eye. The RGB value on his shadowed dark brown shoe is 11, 9, 7, so just barely visible. I chose my lighting intensities to achieve this dynamic range. All my materials' colour values are chosen by how they work when they're sitting outside in the sun next to my reference sphere. The second attachment shows the cab driver next to the calibration image on my monitor. The increased contrast comes from the imager in my cheap phone.

Would you prefer to see the first Netrender image of the cab driver? As you can see, it's histogram is pretty limited as well. Perhaps I should take the monitor calibration image, decal it to a surface and render that?

image_dynamic_range.jpg

image_dynamic_range_compare.jpg

Share this post


Link to post
Share on other sites

These images look generally appropriate on my monitor but still a bit darkish.

The earlier tiffs were so dark that I couldn't discern much in them and couldn't see a difference in the rendering.

Share this post


Link to post
Share on other sites

Here are the two tests I have used with my monitor.

Warning: you can not use these in a browser. You have to load them into a paint program with no color management on and no scaling.

I have the monitor contrast and brightness set so i can distinguish all the bars in colorbands.png

Then i adjusted the monitor gamma so that vertical bands in gamma-test-2.png match each other at 2.2

the original directions can be read here http://www.lagom.nl/lcd-test/  but don't use the test images in a browser.

After i make these adjustments, the graphics and photos i encounter on the web look generally appropriate and full-range.

gamma-test-2.zip

Share this post


Link to post
Share on other sites
Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...