search

OSSOS / MOP

The Moving Object Pipeline: discovery & tracking of trans-Neptunian objects
GNU General Public License v3.0
4 stars 9 forks source link

[cands] chip flip issue not showing sources, making cutouts go backwards #335

Closed mtbannister closed 9 years ago

mtbannister commented 9 years ago

Chips with numbers higher than about 18 are displaying candidate cutouts that a) have no sources in them b) the direction of motion (up and to the left as the three images are stepped through) is opposite its normal down and to the right.

so given it seems to be chip-number related, think it's a cutouts issue. I haven't gone off field L+0+0 yet.

ijiraq commented 9 years ago

This is a bug I introduced…

JJ

On Oct 14, 2014, at 3:37 PM, Michele Bannister notifications@github.com wrote:

Chips with numbers higher than about 18 are displaying candidate cutouts that a) have no sources in them b) the direction of motion (up and to the left as the three images are stepped through) is opposite its normal down and to the right.

so given it seems to be chip-number related, think it's a cutouts issue. I haven't gone off field L+0+0 yet.

— Reply to this email directly or view it on GitHub.

mtbannister commented 9 years ago

I've cleared the 'done' tags on all genuine L+0+0 .p files and deleted the outputs from cands.

ijiraq commented 9 years ago

can you do the same for the ’s’ versions? Were you being shown ’s’ versions?

JJ

On Oct 14, 2014, at 3:53 PM, Michele Bannister notifications@github.com wrote:

I've cleared the 'done' tags on all genuine L+0+0 .p files and deleted the outputs from cands.

— Reply to this email directly or view it on GitHub.

mtbannister commented 9 years ago

I had turned off the randomise in ossos/gui/app.py @property def should_randomize_workunits(self): return False #VERY TEMPORARY FOR REPROCESSING L BLOCK ONLY

so it was presenting files in directory-listing order, which makes the order genuine p for a given field, genuine s for a given field, then after all the genuines are iterated, all the fks.

So I was not being shown 's' versions, hadn't got that far through the field..

mtbannister commented 9 years ago

I think a combination of 001a567c0b61baf986ba70e887fc257b35d28c01 and 1f9e3cefa87f8e8c4488f4a51aadd40f16fa551d have fixed this problem. Has not been apparent where it occurred yesterday in field 0+0 on L-redo.

ijiraq commented 9 years ago

Yes, I forgot to mention in the commit that I resolved this issue. I likely created other issues, since we don’t have unit tests. jj

On Oct 15, 2014, at 9:50 AM, Michele Bannister notifications@github.com wrote:

I think a combination of 001a567 and 1f9e3ce have fixed this problem. Has not been apparent where it occurred yesterday in field 0+0 on L-redo.

— Reply to this email directly or view it on GitHub.