Page 2 of 2 FirstFirst 12
Results 26 to 47 of 47

Thread: PyBot [Deep learning to object detection]

  1. #26
    Join Date
    Nov 2015
    Posts
    16
    Mentioned
    0 Post(s)
    Quoted
    9 Post(s)

    Default

    interesting

  2. #27
    Join Date
    Mar 2016
    Posts
    9
    Mentioned
    0 Post(s)
    Quoted
    2 Post(s)

    Default

    Quote Originally Posted by PshYouLost View Post
    I used a sort of LSTM and just made predictions based on single time intervals until the correct x and y coordinate
    Thanks for your reply Will try to replicate it

    Location tracking based on minimap looks solid

  3. #28
    Join Date
    Mar 2016
    Posts
    9
    Mentioned
    0 Post(s)
    Quoted
    2 Post(s)

    Default

    Quote Originally Posted by Grunt View Post
    Currently testing power mining iron using pybot? 24/7 running no breaks and script have no randoms whatsoever. Now if the theory about using hardware mouse is right?I should not get banned
    For how many days have you been doing that? Still no bans?

  4. #29
    Join Date
    Nov 2006
    Posts
    2,369
    Mentioned
    4 Post(s)
    Quoted
    78 Post(s)

    Default

    I checked your rock training data and it seems you are using bounding boxes as your output. Have you tried using a same size bitmap as output where each rock pixel is colored. This way the neural network would output for each pixel a probability of being part of a rock. In the script you could then choose the pixel with highest probability.

    It would be great if a reflection/injection bot could be used to create those images but otherwise you could just draw a circle or square on top of each rock and convert those images to black and white bitmaps.
    Quote Originally Posted by DeSnob View Post
    ETA's don't exist in SRL like they did in other communities. Want a faster update? Help out with updating, otherwise just gotta wait it out.

  5. #30
    Join Date
    Dec 2008
    Posts
    135
    Mentioned
    0 Post(s)
    Quoted
    44 Post(s)

    Default

    Quote Originally Posted by weequ View Post
    I checked your rock training data and it seems you are using bounding boxes as your output. Have you tried using a same size bitmap as output where each rock pixel is colored. This way the neural network would output for each pixel a probability of being part of a rock. In the script you could then choose the pixel with highest probability.
    If I understand correctly, you are talking about image segmentation. I haven't thought about trying it out because of 2 reasons, the first is the difficulty in collecting training data (maybe your second suggestions can fix that).

    Quote Originally Posted by weequ View Post
    It would be great if a reflection/injection bot could be used to create those images but otherwise you could just draw a circle or square on top of each rock and convert those images to black and white bitmaps.
    Using a bot to create training data is an excellent idea. It will be a very valuable contribution.

  6. #31
    Join Date
    Nov 2006
    Posts
    2,369
    Mentioned
    4 Post(s)
    Quoted
    78 Post(s)

    Default

    Quote Originally Posted by Grunt View Post
    If I understand correctly, you are talking about image segmentation. I haven't thought about trying it out because of 2 reasons, the first is the difficulty in collecting training data (maybe your second suggestions can fix that).


    Using a bot to create training data is an excellent idea. It will be a very valuable contribution.
    I did a google search on that term and it seems to be roughly what I'm talking about: https://leonardoaraujosantos.gitbook...mentation.html.

    I think it might be a more natural task for a neural network but I have never actually tried doing it.
    Quote Originally Posted by DeSnob View Post
    ETA's don't exist in SRL like they did in other communities. Want a faster update? Help out with updating, otherwise just gotta wait it out.

  7. #32
    Join Date
    Dec 2011
    Location
    Toronto, Ontario
    Posts
    6,423
    Mentioned
    84 Post(s)
    Quoted
    863 Post(s)

    Default

    Logged in for the first time in a few months to respond to this.
    This is really great stuff. I was working on a PvM Aid with TensorFlow/OpenCV before I started my job, but this blows it out of the water. Super excited to see where this goes.

  8. #33
    Join Date
    Dec 2008
    Posts
    135
    Mentioned
    0 Post(s)
    Quoted
    44 Post(s)

    Default

    update: finished amethyst bot. Added example script using finite state machine. Generate nice pretty graphs.
    Screenshot from 2017-08-28 03-19-09.jpg

  9. #34
    Join Date
    Mar 2012
    Posts
    107
    Mentioned
    2 Post(s)
    Quoted
    49 Post(s)

    Default

    What are some resources you've used if others want to do something similar but with different activities?

  10. #35
    Join Date
    Dec 2008
    Posts
    135
    Mentioned
    0 Post(s)
    Quoted
    44 Post(s)

    Default

    Quote Originally Posted by One Kid View Post
    What are some resources you've used if others want to do something similar but with different activities?
    opencv
    tensorflow (SSD for detection)
    pytransitions (fsm)
    mss (fast way to capture screen)

  11. #36
    Join Date
    Mar 2006
    Posts
    795
    Mentioned
    0 Post(s)
    Quoted
    6 Post(s)

    Default

    Awesome work, would love to see your code .



    Dont Steal..

  12. #37
    Join Date
    Mar 2012
    Posts
    107
    Mentioned
    2 Post(s)
    Quoted
    49 Post(s)

    Default

    Quote Originally Posted by SubiN View Post
    Awesome work, would love to see your code .
    Agreed! I've been watching this thread closely and would love to see it open-sourced, or just the functioning code.

  13. #38
    Join Date
    Dec 2011
    Location
    East Coast, USA
    Posts
    4,231
    Mentioned
    112 Post(s)
    Quoted
    1869 Post(s)

    Default

    Very cool. Would love to see it on GitHub so I can hack at it.

    Side note: if I was looking to do something like this, where should I start? Learning Python seems like a good jumping-off point.
    Last edited by KeepBotting; 10-18-2017 at 09:25 PM.
    GitLab projects | Simba 1.4 | Find me on IRC or Discord | ScapeRune scripts | Come play bot ScapeRune!

    <BenLand100> we're just in the transitional phase where society reclassifies guns as Bad™ before everyone gets laser pistols

  14. #39
    Join Date
    Mar 2013
    Posts
    32
    Mentioned
    0 Post(s)
    Quoted
    10 Post(s)

    Default

    Quote Originally Posted by legitmail View Post
    For how many days have you been doing that? Still no bans?
    I'm also curious if you can provide any insight into the hardware vs. virtual mouse discussion.

  15. #40
    Join Date
    Jul 2012
    Posts
    33
    Mentioned
    0 Post(s)
    Quoted
    5 Post(s)

    Default

    Any progress? PoC code? Would this be implemented into Simba, a new framework or a combination of programs? I really look forward to any future updates!

  16. #41
    Join Date
    Mar 2018
    Posts
    2
    Mentioned
    0 Post(s)
    Quoted
    0 Post(s)

    Default

    Here, I've made some training data for image segmentation

    villavu.com/forum/showthread.php?t=118169

  17. #42
    Join Date
    Feb 2008
    Posts
    6
    Mentioned
    0 Post(s)
    Quoted
    1 Post(s)

    Default

    Any updates on this project?

  18. #43
    Join Date
    Nov 2014
    Posts
    17
    Mentioned
    0 Post(s)
    Quoted
    5 Post(s)

    Default

    Would love to work on this! Ive been thinking about designing a RL pk bot for a while...

  19. #44
    Join Date
    Jul 2015
    Location
    San Diego
    Posts
    14
    Mentioned
    0 Post(s)
    Quoted
    7 Post(s)

    Default

    Very cool project! I was actually going to embark on something very similar to this using Lazarus/Delphi so that it could be easily integrated with Simba. I'm curious, how taxing is this on your machine/what are the specs on the machines you've tested this on? I'd imagine a decent modern computer would have no problems running a single instance of this, but I'm thinking it might not scale too easily.

  20. #45
    Join Date
    May 2011
    Location
    In an Island.
    Posts
    1,413
    Mentioned
    2 Post(s)
    Quoted
    149 Post(s)

    Default

    Got me quite interested on this since it’s Machine Learning related!
    ''If you want something you've never had, you have to do something you've never done''


    total leve 2715/1b exp +... exterminated.

  21. #46
    Join Date
    Jul 2015
    Posts
    4
    Mentioned
    0 Post(s)
    Quoted
    0 Post(s)

    Default

    Im assuming this has been abandoned but it gave me motivation to learn about ML myself. I am taking a 4** level ML class next semester so a head start wouldn't be too bad

  22. #47
    Join Date
    Mar 2006
    Posts
    795
    Mentioned
    0 Post(s)
    Quoted
    6 Post(s)

    Default

    reviving this thread , found this neat source if anyone else wants to learn
    https://github.com/slyautomation/osrs_yolov5



    Dont Steal..

Page 2 of 2 FirstFirst 12

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •