Closed DanielCoelho112 closed 2 years ago
Hi @miguelriemoliveira,
to give continuity to my desert pass, this is proving to be harder than I thought.
So we have two options to change the illumination:
Services
I create a script that changes the illumination via SetLightProperties service. The service works in a way that if I do SetLIghtProperties, and then GetLightProperties, the information retrieved is the one that I sent. The problem is that in gazebo nothing changes...
Topic
There is a topic called "/gazebo/my_room_024/light/modify" that we can use to change the lighting conditions. When I do:
gz topic -i /gazebo/my_room_024/light/modify
I get this:
Type: gazebo.msgs.Light
The problem is that I cannot import that message... The documentation of that msg is the following: https://osrf-distributions.s3.amazonaws.com/gazebo/msg-api/dev/light_8proto.html
I read somewhere that I need to build that message, but I have no idea how to do it, do you have any ideas? I looked around and I found this:
https://github.com/arpg/Gazebo/blob/master/gazebo/msgs/light.proto
I was thinking about creating our own custom message, but the extension of this file is .proto, and I don't know what to do with it.
Hi @miguelriemoliveira and @pmdjdias,
Now we can produce datasets with variable lighting conditions.
I added 6 lights, check the model with lights almost completely off:
and now with lights completely on:
We have 2 modes:
In this case, in each frame a light intensity is randomly generated. Run:
rosrun localbot_core data_collector --mode automatic_path --seq seq1 -nf 200 -uvl -rl
Using these methods, in consecutive frames, we can obtain a significate light difference, check these images:
In this case, the logic is similar to the poses. We have a continuous variation of the lights over time.
Run:
rosrun localbot_core data_collector --mode automatic_path --seq seq1 -nf 200 -uvl
Nice, looking forward to see the impact on training ;-)!
Yes, I also like it. One light I think is missing is the outside directional light of the sun.
In any case I think this looks very nice: we now have these highly illuminated walls which are actually very realistic, Lucas had to take some real images that looked very much like this from the color correction.
This is done.
To increase the robustness of our models with respect to various lighting conditions throughout the day, we should produce datasets with variable lighting conditions.