Stream Files missing from Feature Extraction window #446
Comments
|
Hi! Thanks for raising this. Indeed it is not yet described well in the documentation and we are hosting only a limited amount of features on github for now. We plan to add additional ones to the interface, but at this time you need to add them manually. Make sure you point your cml directory in the database settings also to the cml folder. If you only downloaded the exe you will have to add the CML folder from here https://github.com/hcmlab/nova/tree/master/bin/cml As you are intending to recognise smiles, i recommend using a featureset for facial features, such as openface. We are currently not hosting this publicly on github but I added a nova-ready-to-use openface2 (https://github.com/TadasBaltrusaitis/OpenFace) component here: https://mediastore.rz.uni-augsburg.de/get/N1IsN68cRx/ Please see Licence instructions in the Openface Git. Make sure you extract the content of this folder so you have the structure nova/cml/chains/video/openface2. In the first screenshot you posted, under "streams" -> edit make sure the files have the type "video" (needs to match the cml/chains/video folder structure). and that the samplerate of the video is correct (e.g 25 or 30 frames per second) Now you should see the video and the openface2 chain. When you extract the features for the first time, it will download all necessary files. It should then create the stream files automatically and add them to the database. (might take some time depending on the length of your video and the performance of your pc). For the training process, make sure you have the the cml/models/templates/.. folder checked out. also note that you will need at least one annotation in the database to train the model. (You can try to start with linsvm which is very fast) Please let me know if this helps, or if you need further assistance. Best, |
|
Thank you very much for this detailed answer. It helped a lot Could you please point at some directions to what to look at to solve this issue? |
|
Hi, does this also happen on the 30hz stream? I could image it may be a problem with the 29.99 and 29.98 streams. (try to add it as 30hz and extract the features again on these). If that is not the problem: Do you get any error before the crash? did you download the cml tools (via ? -> update cml tools) |
|
This issue happens also with the 30hz video. |
|
hi, depending on how long you have been using NOVA you might want to update the cml tools (?-> Update CML tools) and, let the dialog delete the dlls, so they are redownloaded. In case there is no update option, please delete the cmltrain.exe and perform this step again. Please let me know if that helps. |
|
Thank you very much for your feedback. |
|
Dear Dr Baur, I checked the event viewer on windows and I attach the message I get. From my understanding, a folder is missing. Application: nova.exe |
|
Hello, yes Description: The process was terminated due to an unhandled exception. this looks like NOVA tries to create a new folder (probably in cml/models/trainer/....) I can have a look at the code as well, just never ran into this before, so might be system specific. |
|
Thank you very much for your reply. I had tried to run it as admin and it does not work either. The path is not in a restricted folder, I also moved it to another location to make sure that it is not a path issue. |
Hello and thank you for this tool!
I am currently learning how to use the tool so I am using some example videos.
When I am using the annotation ability of the tool, I can see that I have two videos. Please see attached image
However, when I open the feature extraction tab I no longer see these videos
The same when opening the training tab
The two videos are located in a folder /data/name_of_database/name_of_session/ and they are named as role.video_name.mp4
Could you please explain how I should fix this?
The text was updated successfully, but these errors were encountered: