Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    simonecnascimento
    @simonecnascimento
    Hi everyone. How can I create labeled videos from a .csv file with modified coordinates? I know DLC used H5 files to analyze the data, but I am trying to convert csv to H5 with no success.
    2 replies
    Duarte Teixeira Rodrigues
    @Duarte-Rodrigues
    Hello everyone. My question is regarding the DLC2Kinematics. I know that the labels/markers are given according to the pixel-image reference, taking the top left corner as the reference, and so you invert the y axis.
    However, when you calculate the velocity and acceleration, is it still inverted or do you take the cartesian axes as a new reference?
    Austin Pauley
    @austin2118ace:matrix.org
    [m]
    Hey guys, weird problem I came across during the refinement/training process today. My model performs very well for the first ~2700 frames of a 7500 frame video. Then there is a period from around 2700-6200 where tracking for all points is extremely poor (0.01) certainty, and then suddenly at frame 6200 it goes right back to to 0.98-0.99. The behavior during that time period is nothing that hasn't been captured in the initial 100 Frames of training data. I'm planning to do my refinement/training exchange like usual grabbing a lot of frames from this area and training again but just wanted to ask if anyone else has experienced this.
    6 replies
    MukeshNarendran
    @mukeshnarendran7
    Hi , I have annotated data with a model and loaded it in the DLC annotation tool and am able to see my annotations on the GUI but I can not save a few annotations that i wish to make changes. Any hints on what i need to check here. Thanks
    5 replies
    MukeshNarendran
    @mukeshnarendran7
    image.png
    afbrokaw
    @afbrokaw

    Hi All! I'm very new to DeepLabCut and all that is associated (Python, Colab etc). I have been practicing with running the open field demos. I am mostly able to get things to run (starting with the GUI on my machine, then using the Colab notebook to run training, then back to GUI on my machine), but don't run the example for the full recommended iterations (since I'm only testing/learning). After analyzing the video, my plots (such as x pixels versus y pixels) are completely blank. I am still able to create a labeled video showing my labeled points. Can anyone help me figure out what step I might have missed? Thanks in advance.

    Boosting, anyone have advice? I'm pretty sure its something simple, but I'm a coding newb....

    1 reply
    Jorge Maicas-Royo
    @jmaicas
    Hi, I do not seem to find any message talking about an issue I have with maDLC with 2 animals. After the training, the _full.mp4 and the evaluation looks good. However, the tracklets look really bad, big gaps of tracking information was missed in the way. Playing with the .yaml parameters it seemed to improve a bit but not much. And I saw something that never happened to me before. If I create a final video without refining tracklets, it seems to draw either all or none of the bodyparts. When the bodyparts and the skeleton are plotted they look very good but then, some of the body parts are occluded and none of them are plotted, until all of them are visible again. Have anybody any experience with a problem like that? Thanks!
    3 replies
    samaramiller
    @samaramiller
    Hi DLC community, I am having issues with the refine tracklets GUI in maDLC. Pretty sure it is user error on my end and would appreciate any tips. My mouse identities are switched for large chunks of my videos after a mating/fighting bout. What do I need to do after flagging start/stop frames and lassoing animals to make the switch (the "S" hot key doesn't do anything)? Thanks!
    madmaxpython
    @madmaxpython

    Hey DLC community, I have some issues to do the refinement part of my project using maDLC.
    I did the training phase using Google Colab, and now I want to do the refinement of my tracklets using the GUI of DLC, but I have multiple issue:
    First I notice that I don't have the same "tab" (Label Fram, Creat Training Dataset etc...) as most of the tutorials done by Matias's Lab. And in the tab 'OPT: Refine Tracklets', I can choose the config file, the video, but not the tracklet file (.pickle).

    So I tried to do it without, but whenever I use the Flag tool, it give me the following error:
    IndexError: invalid index to scalar variable.

    And If I use the Lasso tool to change the ID of my mice, it give me :
    AttributeError: can't set attribute

    By the way, my computer is a MacBook, and I'm using the DLC version 2.2.0.6
    ajford-23
    @ajford-23
    Hi Everyone, I've been trying to train a single animal model on labeled videos that were 110 by 110 pixels in size. However, DLC v2.2.0.1 doesn't update the terminal after it says "Starting Training....". DLC ended up just being stuck. I even set display iteration to 1, but DLC appears to hang up. I thought it might've been a driver issue, but when I began training on videos that were 1280x1024 px in size, training runs just fine. I was wondering if anyone knows something about a minimum size for videos in DLC or if anyone has encountered similar issues.
    1 reply
    tomwilson41986
    @tomwilson41986
    image.png
    Hi, i keep getting an error when trying to add new videos to an existing model. Any tips?
    1 reply
    rhgoo
    @rhgoo
    Hi DLC community, I'm trying to understand what the specific calculations are for 'likelihood' after evaluating and analyzing videos w/ a network. How is likelihood calculated and what possible conclusions can be derived from it?
    1 reply
    austin2118ace
    @austin2118ace

    Hi Guys,

    I'm having some trouble with my model and processing videos. We have used 4 videos at about 5 minutes each, and extracted around 250 frames total. I use one video at a time, train for ~100k iterations, process a second video, and use refine outliers to add a new video and start a new iteration. I got to iteration four, and the model was performing fairly well. I tried to process one of our experimental videos. These videos are ~40 minutes long, but are identical in lighting, experimental hardware, and animals as the training videos.

    We usually measure performance with what percentage of frames are under 0.95 certainty. We are shooting for ~5% of frames under 0.95 and will smooth out the other 5% using some post-processing. At the end of iteration 4 we processed a experimental video and got very good certainty numbers; however, all of the actual XY coordinates were wrong. The points were occasionally placed in the correct place, but often they were just flickering around the video. It was strange because the model was very certain they were all correct, and performance while we were training was generally good. I then processed a different experimental video and performance is very poor. With over 50% of the frames showing uncertainty <0.95 across all our labeled points.

    The later leads me to believe there are some subtle changes that the model isn't trained for, but what about the former video where the model was very certain but mislabeled everything? Would the mismatch in video length/size have any affect on the ability to properly label points? Should my training videos be longer to get a wider variety of subtle changes in lighting and animal movement? I feel like I captured most of the behavior in my labeled frames very well. So I'm not sure where the problem lies.

    mildasthma23123214142
    @mildasthma23123214142:matrix.org
    [m]
    Hi
    is anyone willing to help me with this assignment I have using DeepLabCut model that I imported into Colab
    Muad Abd El Hay
    @Cumol
    I have a machine with 1024 cores and 4000gb of RAM. What is the most efficient way to run DLC using that machine?
    1 reply
    bfm222
    @bfm222
    Hi all, in the Create Videos step, when I select yes for "create video with animal ID colored", the resulting video often switches our two fish colors when they are near each other. Does anyone know how I can mitigate this?
    1 reply
    NewmanAHuman
    @NewmanAHuman
    Just read the new maDLC pub - is there an update to dlc-live (tf2-enabled?) for real-time multi-animal pose estimation? (Congrats on pub, btw!:)
    M
    @mchaker
    multi-GPU question: does DLC actually use multiple GPUs by default? I'm running the Docker_TrainNetwork_VideoAnalysis.ipynb notebook and it seems to only be using CUDA device id 0 (I have 8 devices). In the logs for the deeplabcut.train_network() call, all 8 devices are displayed, but the work seems to only be going to device 0.
    2 replies
    @MMathisLab
    tomwilson41986
    @tomwilson41986
    image.png
    Just started getting this error in Colab. Anyone seen it before?
    1 reply
    Raineer Mozo
    @mozoraineer
    Hello, anybody who encountered this weird error after the 2.2.1 release? i suddenly get this error. Tried clean reinstalling my anaconda and DLC but still got this error. I found a solution at the github but im just asking if im just missing something simple. image.png
    2 replies
    Ritu Panda
    @rp13:matrix.org
    [m]
    How to label the extracted frames from the video without using the GUI
    1 reply
    Muad Abd El Hay
    @Cumol

    Is there an issue with using multiple GPUs? For some reason, using 4 instead of 1 (of the same) GPU only leads to an minor increase in inference speed (from 75fps to 85 fps), but that doesn't make much sense...

    Additionally, changing batchsize in the analyze_videos call does not lead to faster inference either.

    Am I missing something?

    7 replies
    stickzh
    @stickzh
    Has there been some changes in the ast month?(I expect so)
    on colab I was using:
    !pip install statsmodels==0.13.1
    !pip install deeplabcut
    !pip install opencv-python-headless==4.5.1.48
    !pip install --upgrade tables
    !pip install tensorflow==2.8
    but now import deeplabcut is faiing with:
    "AttributeError: module 'matplotlib.cbook' has no attribute '_make_class_factory'"
    2 replies
    KaylaPitts
    @KaylaPitts
    image.png
    image.png
    Received an error message while trying to save labeled images. Its about the tables not being installed, but it appears that it is and fully upgraded. Any suggestions?
    1 reply
    JHeroine
    @JHeroine
    Hi, Guys!
    Does anyone here know why the same label dot is two in a test image in the evaluation folder?
    One is for the train result dot and the other is for the test result..?
    2 replies
    Muad Abd El Hay
    @Cumol
    For some reason, DLC works (and produces files) with some encodings but not with others. Anyone has a clue if this is a general bug or something I can fix on my side?
    2 replies
    jkopp
    @Tetra-quark

    Does anyone know how the mirror parameter is used in the train/pose_cfg.yaml?

    I found DeepLabCut/DeepLabCut#939:

    "Mirror is used when you have two symmetrical points you want treated the same during training, like two knees."

    This is certainly not what I want and it may be true for imgaug, but unless I have misunderstood something I am sure this is not what the deterministic augmenter does.

    The deterministic augmenter seems to handle mirror augmentations in the way that I want: transforming the coordinates of the joints (flip left to right) and then swapping the joint names, only for the joints that are chiral (ie have a symmetric counterpart). The symmetric association of the chiral labels is specified by the list of lists in the all_joints parameter in the pose_cfg.

    So by using mirror: true I expect to effectively double my training data and reduce left/right asymmetry in the predictions of my model.

    Can anyone confirm this?

    afbrokaw
    @afbrokaw
    Hello! I am testing out DeepLabCut on some old data from our lab, and am having trouble getting the diagnostic plots to show up. I am referring to the optional plots that can be created after the network is evaluated. Where in the project folder do those get stored? The 'plot poses' folder that is created within the video folder is just empty. Or sometimes, the plots are created, but no data is actually shown on them - its just an empty plot. Can someone advice what I am missing? Is it something going wrong with my model training? Or just an error with the plotting? I had asked about this before, but never got a response, and I did get it to work once on the DLC open field demo data (though sometimes even that didn't plot properly). Thanks in advance.
    4 replies
    CorinneAKelly
    @CorinneAKelly
    If I have labels from one project, is there a way to add those same labels to a different project without relabeling the frames?
    2 replies
    Khiem Vuong
    @kvuong2711
    Hello! I see that DeepLabCut has a web-based GUI for labeling here (https://contrib.deeplabcut.org/#). Is there a plan to release the source code for this? My simple goal is to deploy a web-based GUI for labeling, e.g., hosting a web-based GUI so that the annotators don't have to install anything (just need to go to the website and annotate). Is there a simple way to make this work? Thank you!
    kaya75
    @kaya75:matrix.org
    [m]
    Hello everyone, I compressed my video before using the 'label_frames' function, because it was too heavy. But when it open it on GUI, the video lost a lot of its resolution, and it's impossible for me to see some important things as mouse's legs. Do you know how to improve the quality of the video during labeling task ? Thanks !!
    2 replies
    KennyT
    @KennyTC
    Hello, does anyone know that
    given input is an image, how can I get the coordinates of each prediction points? I want to use those coordinates to calculate many other things.
    2 replies
    LyDNeuRoN
    @LyDNeuRoN
    hello everyone, I have a problem when I try to label my video. Whenever I click on the part I want to label no color dot appears. I don't know what is happening since the terminal seems to be fine.
    7 replies
    kaya75
    @kaya75:matrix.org
    [m]
    I think there is a small box near the choice of the size of the dots nope ? You have to click on it first I guess ..
    Raineer Mozo
    @mozoraineer
    Hi, is there any way to make the model infer realtime? without using dlc live? any inputs would be greatly appreciated
    1 reply
    kaylago
    @kaylago
    I extracted outlier images from analyzed videos in Colab before realizing that doing so in Colab would result in errors - the images extracted in Colab have no labels on them to refine. I would like to simply delete the outlier frames and their corresponding rows from the machine h5 and machine csv files, however I’m not sure what function to use to do this. The function deeplabcut.dropannotationfileentriesduetodeletedimages(config) does not work in this case…Is there an alternate function I could use? Thanks!!
    LyDNeuRoN
    @LyDNeuRoN
    Hello everyone, I have a problem when I try to Train the network. Every single time on the terminal appears: "Allocation of 104693760 exceeds 10% of free system memory". And it will stay like this for hours .... It is stuck :( Anyone can help ??
    2 replies
    LoganFriedrich
    @LoganFriedrich
    Does anyone know how to access a and use a model for a project created with the GUI without the GUI? For instance, I'm trying to access a project called "MSA - V2-LF-20220303" and tried to write a script to do so. There's a little more to it as I also have some file movements with shutil included, but here's the basics:

    Analyze videos in A:!!! DLC Input with deeplabcut

    def getsubfolders(folder):
    ''' returns list of subfolders '''
    return [os.path.join(folder,p) for p in os.listdir(folder) if os.path.isdir(os.path.join(folder,p))]

    def do_analysis():
    project=r"MSA - V2-LF-20220303"
    shuffle=1
    prefix=r"A:\AIs"
    projectpath=os.path.join(prefix,project)
    config=os.path.join(projectpath,"config.yaml")
    basepath=r"A:!!! DLC Input"
    subfolders=getsubfolders(basepath)
    for subfolder in subfolders: #this would be January, February etc. in the upper example
    print("Starting analyze data in:", subfolder)
    subsubfolders=getsubfolders(subfolder)
    for subsubfolder in subsubfolders: #this would be Febuary1, etc. in the upper example...
    print("Starting analyze data in:", subsubfolder)
    for vtype in [".mp4",".m4v",".mpg"]:
    deeplabcut.analyze_videos(config,[subsubfolder],shuffle=shuffle,videotype=vtype,save_as_csv=True)

    Time to analyze files

    schedule.every().day.at("16:15:00").do(do_analysis)

    bfm222
    @bfm222
    Screenshot from 2022-05-23 17-09-57.png
    Hi all, I'm having trouble with the Analysis step of DLC. The four .pickle files for a given video are created, but the .h5 file isn't. I'm getting a RecursionError (maximum recursion exceeded while calling a Python object) and a ValueError (could not reconstruct 2 tracks from the tracklets given).
    I'd appreciate any help or advice!
    LuckyyyClover9
    @LuckyyyClover9
    Hello! We ran into an issue where h5 files weren't formatting from labeling frames due to an error with the tables package. We have csv files from labeled frames, is there any way to convert them to h5 files?