Hello everyone,
i have 3 questions first, i am trying to do a box around the object in the field whether it is a car , human, or whatever I did a script i can capture a few images but when i activate bounding box 2d it is a file format json and npy . any solution? the second question is there a possibility to detect human posing? third, if I have multiple cameras can i record videos simultaneously
Hi there,
for 1: you can use the bounding box data to append boxes to the images, this is not done by default since it will ruin your training data. See the plot_boxes example function in /omni.isaac.synthetic_utils/omni/isaac/synthetic_utils/visualization.py which uses matplotlib.pyplot for this.
for 2: you can use the skeleton_data annotator:
self.annotators.append(AnnotatorRegistry.get_annotator("skeleton_data"))
for 3: yes, if by videos you mean synthetic data recording: here is a snippet including this:
https://docs.omniverse.nvidia.com/app_isaacsim/app_isaacsim/reference_python_snippets.html#multi-camera
@ahaidu Thank you for helping, but the snippet is only for capturing images only I can’t record all cameras at once, i have about 8 cameras and I want to record video for every single camera at the same time
regarding question 1: it solved for me half a way to get the boundary box around the object but is it possible to capture an image bounding2dbox and save it as jpg format, not in npy format?
Multiple cameras should work without any issues, can you provide what problems you are facing with this?
Regarding saving to jpg you will need to write a custom data exporter for that, you can use the example snippet from 3, either to create your own writer or by getting the data directly from the annotator.
@ahaidu The problem I am facing now is that I can only record one video per simulation from one camera using movie capture. My aim that I have 4 cameras on a vehicle and I need to capture 4 recorded videos at the same time so that when I open my directory folder I can find 4 mp4 files but the extension of movie capture can only record video using one camera path
Replicator currently saves the data as images and not as movie files, this might be supported in future releases. You could however easily stich together he .png files using ffmpg after the recording, or write a custom writer which does this on the fly using a python module such as ffmpeg with python bindings.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.