Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Jan 23 16:00
    beniz synchronize #677
  • Jan 23 16:00

    beniz on rotate_action

    added image rotation action, us… added save_path option to speci… added relative/absolute orienta… and 1 more (compare)

  • Jan 23 12:27
    beniz labeled #694
  • Jan 23 11:03
    beniz labeled #694
  • Jan 23 11:03
    beniz assigned #694
  • Jan 23 11:03
    beniz labeled #694
  • Jan 23 11:03
    beniz opened #694
  • Jan 23 10:57

    beniz on ocr_conf

    OCR/CTC confidence output on pr… (compare)

  • Jan 22 13:36
    fantes edited #688
  • Jan 22 13:27
    fantes synchronize #688
  • Jan 22 10:06

    beniz on master

    docker build files update Merge pull request #684 from jo… (compare)

  • Jan 22 10:06
    beniz closed #684
  • Jan 22 06:44
    beniz commented #689
  • Jan 22 06:44
    beniz synchronize #693
  • Jan 22 06:44

    beniz on chain_action_logger

    fixed skipping bad boxes when c… (compare)

  • Jan 21 15:35
    YaYaB commented #689
  • Jan 21 15:25
    beniz labeled #693
  • Jan 21 15:24
    beniz commented #693
  • Jan 21 13:24
    beniz commented #689
  • Jan 21 13:22
    beniz labeled #693
Emmanuel Benazera
@beniz
see my comment, refinedet models cannot be converted to TRT.
YaYaB
@YaYaB
Yes, but you added "Now, you can remove objectness_score from all your models, it does not affect the results at test time." ===> that's why I was confused
It would be nice to have this somewhere in the docs ===> list of models or architectures usable with TRT backend.
Romain Guilmont
@rguilmont
Hello guys, i'm playing with the nice chain API now, but i have some trouble with concurrent requests. When i try to do concurrent chain requests on the same URL ( i.g. /chain/yolo ) i get several 500 errors, but when i randomize these URL, i get none.
This message was deleted

See the logs bellow :

./main --url http://localhost:8080 --batchsize 1 -concurrency 5            
2020/01/09 16:23:27 Running tester for chains with batchsize : 1 - real size of batch 1
2020/01/09 16:23:27 Starting new thread of requests...
2020/01/09 16:23:27 Starting new thread of requests...
2020/01/09 16:23:27 Starting new thread of requests...
2020/01/09 16:23:27 Starting new thread of requests...
2020/01/09 16:23:27 Starting new thread of requests...
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {200 OK} 649 <nil>
http://localhost:8080/chain/yolo {200 OK} 431 <nil>
http://localhost:8080/chain/yolo {200 OK} 409 <nil>
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {200 OK} 1088 <nil>
http://localhost:8080/chain/yolo {200 OK} 1031 <nil>
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {200 OK} 474 <nil>
http://localhost:8080/chain/yolo {200 OK} 624 <nil>
http://localhost:8080/chain/yolo {200 OK} 749 <nil>
http://localhost:8080/chain/yolo {200 OK} 668 <nil>
http://localhost:8080/chain/yolo {200 OK} 888 <nil>
http://localhost:8080/chain/yolo {200 OK} 924 <nil>
http://localhost:8080/chain/yolo {200 OK} 926 <nil>
http://localhost:8080/chain/yolo {200 OK} 967 <nil>
http://localhost:8080/chain/yolo {200 OK} 917 <nil>
http://localhost:8080/chain/yolo {200 OK} 837 <nil>
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {200 OK} 934 <nil>
http://localhost:8080/chain/yolo {500 InternalError} 0 DeepDetect Internal Server Error
http://localhost:8080/chain/yolo {200 OK} 280 <nil>

With randomized URL :

./main --url http://localhost:8080 --batchsize 1 -concurrency 5 --randomURL
2020/01/09 16:24:15 Running tester for chains with batchsize : 1 - real size of batch 1
2020/01/09 16:24:15 Starting new thread of requests...
2020/01/09 16:24:15 Starting new thread of requests...
2020/01/09 16:24:15 Starting new thread of requests...
2020/01/09 16:24:15 Starting new thread of requests...
2020/01/09 16:24:15 Starting new thread of requests...
http://localhost:8080/chain/yolo_10667007354186551956 {200 OK} 567 <nil>
http://localhost:8080/chain/yolo_6334824724549167320 {200 OK} 720 <nil>
http://localhost:8080/chain/yolo_11998794077335055257 {200 OK} 742 <nil>
http://localhost:8080/chain/yolo_13260572831089785859 {200 OK} 874 <nil>
http://localhost:8080/chain/yolo_8674665223082153551 {200 OK} 1080 <nil>
http://localhost:8080/chain/yolo_7504504064263669287 {200 OK} 524 <nil>
http://localhost:8080/chain/yolo_3510942875414458836 {200 OK} 703 <nil>
http://localhost:8080/chain/yolo_4324745483838182873 {200 OK} 886 <nil>
http://localhost:8080/chain/yolo_11926759511765359899 {200 OK} 989 <nil>
http://localhost:8080/chain/yolo_11239168150708129139 {200 OK} 767 <nil>
http://localhost:8080/chain/yolo_3328451335138149956 {200 OK} 946 <nil>
http://localhost:8080/chain/yolo_7955079406183515637 {200 OK} 870 <nil>
http://localhost:8080/chain/yolo_2740103009342231109 {200 OK} 737 <nil>
http://localhost:8080/chain/yolo_1905388747193831650 {200 OK} 833 <nil>
http://localhost:8080/chain/yolo_15649472107743074779 {200 OK} 872 <nil>
http://localhost:8080/chain/yolo_261049867304784443 {200 OK} 798 <nil>
http://localhost:8080/chain/yolo_5600924393587988459 {200 OK} 847 <nil>
http://localhost:8080/chain/yolo_9956202364908137547 {200 OK} 806 <nil>
On deepdetect side, nothing really interesting, this is what i get :
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.499] [api] [info] 127.0.0.1 "POST /chain/yolo" 200 199
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.541] [yolo] [info] number of calls=3
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.541] [yolo] [info] [0] / executing predict on service face_detection
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.669] [yolo] [info] [1] / executing action crop
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.670] [yolo] [info] [2] / executing predict on service age
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.696] [api] [info] 127.0.0.1 "POST /chain/yolo" 200 322
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.881] [yolo] [info] [0] / no result from prediction
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.881] [api] [info] 127.0.0.1 "POST /chain/yolo" 200 340
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.926] [yolo] [info] number of calls=3
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.926] [yolo] [info] [0] / executing predict on service face_detection
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.928] [api] [error] 127.0.0.1 "POST /chain/yolo" 500 0
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.968] [yolo] [info] number of calls=3
Jan 09 15:08:07 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:07.968] [yolo] [info] [0] / executing predict on service face_detection
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.104] [yolo] [info] [1] / executing action crop
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.105] [yolo] [info] [2] / executing predict on service age
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.141] [api] [info] 127.0.0.1 "POST /chain/yolo" 200 215
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.281] [yolo] [info] [1] / executing action crop
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.281] [yolo] [info] [2] / executing predict on service age
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.295] [api] [info] 127.0.0.1 "POST /chain/yolo" 200 326
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.297] [yolo] [info] number of calls=3
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.297] [yolo] [info] [0] / executing predict on service face_detection
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.335] [api] [error] 127.0.0.1 "POST /chain/yolo" 500 0
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.376] [yolo] [info] number of calls=3
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.376] [yolo] [info] [0] / executing predict on service face_detection
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.466] [yolo] [info] [1] / executing action crop
Jan 09 15:08:08 ip-172-31-36-122 deepdetect[2187]: [2020-01-09 15:08:08.466] [yolo] [info] [2] / executing predict on service age
Romain Guilmont
@rguilmont
So my question : is this behaviour expected ? If yes, do you recommand that i randomize the URLs or something else ?
Emmanuel Benazera
@beniz
Hello, the chain API uses a PUT to define a temporary resource. You get an error if you try to PUT it again before it is release. It should be a 500 maybe. For concurrent calls you want to randomize or number the resource as you are doing, this is correct.
Romain Guilmont
@rguilmont
thanks ! Then that's what we'll do.
YaYaB
@YaYaB
Hey guys, I played a bit with the chain service and I saw a bug when bounding boxes obtained are out of range of the image. jolibrain/deepdetect#689
Emmanuel Benazera
@beniz
@YaYaB thanks, will check.
cchadowitz-pf
@cchadowitz-pf
hi @beniz, just checking in - any luck with #687?
Emmanuel Benazera
@beniz
@cchadowitz-pf unfortunately no bandwith here to dig into TF internals as we're all pretty busy this month. Do you believe you could isolate the piece of code (logging within TF, probably with glog) that breaks the build ? Thanks, and apologies for the late answer.
cchadowitz-pf
@cchadowitz-pf
no worries, i understand. I'll see what I can do :)
Emmanuel Benazera
@beniz
sounds good.
jassem123
@jassem123
image.png
Hello guys , I tried building deepdetect on Ubuntu 18.0.4 bionic beaver from source , I ran through a compilation error during cmake that I couldn't find a workaround for. Here's a screenshot , I'd be available to provide any further logs . Thanks in advance
Emmanuel Benazera
@beniz
Hi @jassem123 most likely you are trying to build DD + Tensorflow on a machine that doesn't have enough memory for this build.
Unless you can add memory to the machine / VM, you may want to use Docker images instead.
jassem123
@jassem123
Thank you for your short reply I'll keep working on it
jassem123
@jassem123
Hello again , Any walkthrough for uninstalling the unnecessary dependencies and packages I previously installed to build DD + Tensorflow ? any instructions are much appreciated too
thanks in advance
Emmanuel Benazera
@beniz
I guess you can use apt to uninstall any unwanted package...
jassem123
@jassem123
Do you have any file listing the related packages ? please ?
Emmanuel Benazera
@beniz
Look at the quickstart section on the website, for the builds from source, sure.
jassem123
@jassem123
I wasn't paying enough attention when I started building , on the quickstart page it says that ubuntu's version is 16.0.4 , can't this be the reason behind the cmake failure I have encountered ? thought this was worthy asking before I started deleting xD
Emmanuel Benazera
@beniz
no, the failure you reported is due to the kernel killing the build job. Check how much RAM your machine has.
jassem123
@jassem123
it has 6 out of 8
I can't allocatemore than that
Emmanuel Benazera
@beniz
a build without tensorflow support will work without issue.
jassem123
@jassem123
I have some questions regarding the Docker alternative you have proposed sir , (its concept is new to me )
Can a Docker run just locally ?
I am using deepdetect to classify medias on elasticsearch
as far as i understood
the confidentiality of the indexed media is crucial for my corporate
Emmanuel Benazera
@beniz
Please review Docker documentation and tutorials on the Web.
jassem123
@jassem123
will do , thanks again you've been of great help
Emmanuel Benazera
@beniz
np, good luck
jassem123
@jassem123
Hi man
this is out of context , but am starting to bang my head against the wall in a bit
i ran apt-get autoremove
now ifconfig only shows lo
as loopback
jassem123
@jassem123
Hello
Does this link https://hub.docker.com/r/beniz/deepdetect_cpu has tensorflow integrated ?
Emmanuel Benazera
@beniz
no, deepdetect_cpu_tf has
jassem123
@jassem123
Thanks
does cpu_tf has also Caffe ?
Emmanuel Benazera
@beniz
it does
jassem123
@jassem123
Thank you