I'm trying out the Model Predictive Control script provided in the docs. I notice that when setting enable_visualization=True when instantiating a CausalWorld environment, the real-time visualization of the CEM is very high resolution, which is great. However, the recording of the CEM obtained via gym.wrappers.monitoring.video_recorder.VideoRecorder has much poorer quality.
Is there a way to record the videos such that their resolution is as good as the real-time visualization, or at least better than what we currently obtain from the VideoRecorder?
I suspect that one would have to modify the render() function of CausalWorld to achieve this, but I don't know what exactly needs to be done.
I'm trying out the Model Predictive Control script provided in the docs. I notice that when setting enable_visualization=True when instantiating a CausalWorld environment, the real-time visualization of the CEM is very high resolution, which is great. However, the recording of the CEM obtained via gym.wrappers.monitoring.video_recorder.VideoRecorder has much poorer quality.
Is there a way to record the videos such that their resolution is as good as the real-time visualization, or at least better than what we currently obtain from the VideoRecorder?
I suspect that one would have to modify the render() function of CausalWorld to achieve this, but I don't know what exactly needs to be done.