*** tpb has joined #timvideos | 00:00 | |
*** Bertl_oO is now known as Bertl_zZ | 00:23 | |
*** nueces_ has quit IRC | 01:56 | |
*** nueces_ has joined #timvideos | 02:53 | |
*** nueces_ has quit IRC | 03:01 | |
xfxf | mithro: i'm guessing you recorded the video in this talk, i can see a disassembled opsis at the start? https://news.ycombinator.com/item?id=12528206 | 03:10 |
---|---|---|
tpb | Title: Gorgonia: a library like Theano or TensorFlow, mainly written in Go | Hacker News (at news.ycombinator.com) | 03:10 |
xfxf | also the 15 fps thing was a firmware bug i had ages back, i thought that was fixed? | 03:10 |
mithro | xfxf: That was running the shipping firmware, didn't have time to flash with something newer | 03:37 |
xfxf | ah, that explains the 15fps thing | 03:39 |
xfxf | also,the lack of artifacts - the shipping firmwares doesn't suffer from that? | 03:40 |
xfxf | if so, hopefully that makes tracking down the actual cause easier | 03:41 |
mithro | xfxf: It does | 03:41 |
mithro | xfxf: It does suffer from the artifacts, if I had multiple outputs enabled it was worse | 03:42 |
xfxf | hmm, I'm sure with the firmware we used at pycon, even two inputs, encoder, both outputs off was way worse than what I see in that vid | 03:47 |
xfxf | but I'd need to check | 03:47 |
*** paradisaeidae_ has joined #timvideos | 04:06 | |
xfxf | oh hang, i was testing it with one output enabled too. that'd explain the difference | 04:08 |
CarlFK | start of what? | 04:09 |
xfxf | video linked from the page above | 04:26 |
*** rohitksingh_work has joined #timvideos | 04:33 | |
*** Bertl_zZ is now known as Bertl | 05:06 | |
mithro | CarlFK: http://shop.duxtel.com.au/product_info.php?products_id=355 | 05:52 |
tpb | Title: Mikrotik RouterBoard - Fully Assembled Systems - online catalogue for MIKROTIK products - - CRS210-8G-2S+IN: 8 port Desktop Switch with 10gbps fibre support (at shop.duxtel.com.au) | 05:52 |
CarlFK | xfxf: I see no links to videos on https://news.ycombinator.com/item?id=12528206 | 05:54 |
tpb | Title: Gorgonia: a library like Theano or TensorFlow, mainly written in Go | Hacker News (at news.ycombinator.com) | 05:54 |
*** cr1901_modern has quit IRC | 05:54 | |
xfxf | CarlFK: click on the actual story, it's in there? | 05:56 |
CarlFK | mithro: neat. | 05:56 |
CarlFK | xfxf: heck if I know | 05:56 |
CarlFK | im sure if I follow enough links I will come to something ;) | 05:57 |
mithro | CarlFK: You can get yourself a 10G network card | 05:57 |
xfxf | mithro: have you actually used that switch? | 05:57 |
xfxf | mithro: microtik gear can be.... interesting | 05:57 |
mithro | Nope, but it has evil_steve recommendation | 05:57 |
xfxf | he's using microtik in production at universities? that genuinely surprises me | 05:57 |
mithro | xfxf: Probably not, but for the price point he said it was probably an okay buy | 05:58 |
CarlFK | do we need 10G? (or is this for my 2017 UN thing?) | 05:59 |
xfxf | i've decided against buying any 10G gear now, i'm going to experiment with LACP bonded gigabit first | 06:00 |
xfxf | i doubt i'd ever need more than 2-3G, and it's always going to/from seperate machines, so mac based lacp should in theory be perfectly fine | 06:00 |
xfxf | CarlFK: i think he's talking if we want more than 2-3 uncompressed audio/video ingests | 06:01 |
CarlFK | ah, yeah, I don't need that any time soon. I hope. | 06:02 |
mithro | xfxf: When we get Opsis Ethernet working, that is going to be ~800Mbps | 06:04 |
xfxf | mithro: why would uncompressed video over the Opsis ethernet be more than now? | 06:04 |
mithro | xfxf: I'm unsure why the current value is so low | 06:04 |
CarlFK | mithro: what makes you think it will be ~800Mbps? | 06:05 |
xfxf | mithro: isn't the current rate higher than what it could be, due to us doing full 24-bit color depth or whatever? | 06:06 |
mithro | CarlFK: actual calculations | 06:06 |
CarlFK | mithro: show your work ;) | 06:08 |
CarlFK | except I want to compare it to what gs is doing, and that's probably not something I should try | 06:09 |
rohitksingh_work | mithro: ping | 06:21 |
mithro | rohitksingh_work: pong | 06:21 |
*** Sajuuk has quit IRC | 06:23 | |
*** paradisaeidae_ has quit IRC | 06:28 | |
*** Bertl is now known as Bertl_oO | 06:30 | |
CarlFK | mithro: do you expect the Opsis to be able to connect to vocto core, or would we run a gst pipeline to hook the two together? | 06:33 |
mithro | CarlFK: Probably a pipeline unless there are some changes to voctocore to support it | 06:34 |
mithro | The pipeline should be pretty simple though | 06:37 |
CarlFK | mithro: that also means the pipeline can decompress | 07:04 |
mithro | CarlFK: The frames from the Opsis are uncompressed | 07:04 |
CarlFK | mithro: I am guessing the mjpeg compression could be applied | 07:06 |
mithro | Sure I guess | 07:06 |
mithro | We should be able to just do two streams at 30fps | 07:06 |
CarlFK | mithro: like one for each input? | 07:07 |
mithro | CarlFK: yes | 07:07 |
CarlFK | neat | 07:07 |
mithro | Oh - that is probably the reason - 30fps verse 60fps | 07:08 |
*** automatical has quit IRC | 08:12 | |
*** automatical has joined #timvideos | 08:12 | |
*** rohitksingh_work has quit IRC | 08:25 | |
*** rohitksingh_work has joined #timvideos | 08:31 | |
*** cr1901_modern has joined #timvideos | 09:53 | |
*** CarlFK has quit IRC | 10:00 | |
*** nueces has joined #timvideos | 13:23 | |
*** rohitksingh_work has quit IRC | 13:33 | |
*** rohitksingh has joined #timvideos | 14:21 | |
*** rohitksingh has quit IRC | 14:26 | |
*** rohitksingh has joined #timvideos | 14:27 | |
*** CarlFK has joined #timvideos | 15:24 | |
*** ChanServ sets mode: +v CarlFK | 15:24 | |
*** rohitksingh has quit IRC | 16:25 | |
*** rohitksingh has joined #timvideos | 16:26 | |
*** rohitksingh has quit IRC | 17:45 | |
*** thaytan has quit IRC | 18:52 | |
*** andi-m is now known as saerdnaer | 21:43 | |
*** saerdnaer is now known as andi-m | 21:44 | |
*** puck has quit IRC | 22:03 | |
*** puck`` has joined #timvideos | 22:24 |
Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!