New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
How to apply mask_rcnn segmentation on a balloon video ? #101
Comments
You can find the following config in the tutorial, which is used to train the balloon model:
the config file you're running demo.py with does not have these changes, so it won't work. The model you trained will be in |
I tried to do this but didn't work. Is this because ballon_dataset was directed to /dev/null in the tutorial and somehow this is messing with the paths?
Error
|
|
Similarly, all the changes you made to |
Ok now it is running but the output is the same as input with no segmentation at all.
|
Which means you need to change the config file given to |
to something like this? I ended up with the following error.
Error
|
You need to create a yaml file which has the same content as |
I think that is what I did. I am writing all the steps just to be clear.
configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml looks like the following ..
|
What the command run is
that you made during training, because these changes are not in You need to add these changes either to |
This time I forked Detectron2 repo and re-ran all the code by adding config changes to the but still have issues.
|
As the error says, your code is trying to read a file that does not exist. Since your command has "cd detectron2_repo", you should remove "detectron2_repo" from the path. |
Thank you now having the following issue
|
Set |
Thank you. Seems to be running but got stuck here
|
This error is addressed by 22e04d1 |
If I have to add this new change do I need to repeat the forking/install and rerun things? |
If you install detectron2 in the same way as the tutorial, then |
All ran well but the output is the same as the input. There was no object detection in the video. Do you think this could be a threshold score issue? When I change it from 0.7 to 0.6, there were boxes popping everywhere but not where the balloons are. This is strange.
|
As I said in #101 (comment)
You load a wrong model. |
In the tutorial .pkl file was set to weights. Setting .pth is the same as .pkl? |
Both file extensions are supported.
I do not know what you mean by "when I build the model", and when saying "the .pkl file" it is unclear to me what pkl file you're referring to. |
Here is the line from the tutorial where *.pkl file set to model weights in the demo.py
If I understand correctly, I just replace this .pkl with my own .pth file in demo.py? |
Yes |
Yey!!! it is working. Thank you for being so patient with me. |
I trained just like the balloon tutorial, got the trained result, and I have the same error when predict in my own new dataset, which doesn't have label, only image, I think it's because I don't set the config correctly, would you please tell me how to set the cfg correctly when predict in new dataset. cfg.MODEL.WEIGHTS = os.path.join(cfg.OUTPUT_DIR, "model_final.pth") in () /content/detectron2_repo/detectron2/engine/defaults.py in init(self, cfg) IndexError: tuple index out of range` anyhelp would be appreciate, thank you so much! |
thank you, I figured it out. I need to register the test data before using it, and then do as you say. thanks a lot~ |
Hi, I am going through the google colab example tutorial.
I am trying to apply mask_rcnn segmentation on a random youtube balloon-video instead of an balloon-image to detect balloon only (one class).
How can I assign .yaml and .pkl files that were generated using images earlier in the tutorial to a random video? thanks
I tried the foolowing but it didn't work. I think I am having trouble assign the trained config and model files.
The text was updated successfully, but these errors were encountered: