search

qianqianwang68 / omnimotion

Apache License 2.0
2.13k stars 126 forks source link

Particle Tracking Results #46

Open sathya-muru opened 11 months ago

sathya-muru commented 11 months ago

Hi,

I have been running some test with omnimotion for a microgravity research project. Specifically we are interested in its ability to track falling particles. I took a sample video of larger balls falling to see how omnimotion performs and the results are not great. I tried the video slowed down and a regular speed. Do you know of any configuration changes that could produce better results for this type of video? I used the default configuration file.

https://drive.google.com/file/d/1yx-oDuu_4quwm0bf9t6tfwVE588TjItV/view?usp=sharing https://drive.google.com/file/d/1WRzMdiCTwDibMqbDTsDdFDc4OHZQ6KHc/view?usp=sharing

Rahul-Singhal-Crafty-Apes-VFX commented 11 months ago

Curious if you tried a patterned ball, e.g. soccer ball and then started the video with the ball stationary and in view. Also, does your research project give you flexibility on the frame rate for the camera to go to 60fps or even higher if you have the hardware?

sathya-muru commented 11 months ago

Thanks for your reply! I didn't try a patterned ball, that might be my next experiment. The ball was visible stationary in the first few frames of the video. I extracted images at 55 frame per second, to reduce the amount of disk space required for preprocessing files.

rozumden commented 10 months ago

Hi,

We've been working on a similar research problem for years now. You can check our C++ demo on fast-moving objects detection and tracking: https://github.com/rozumden/fmo-cpp-demo We also have methods to deblur such objects: https://github.com/rozumden/DeFMO and reconstruct their shape and motion in 3D: https://github.com/rozumden/ShapeFromBlur We also developed a learned FMODetect method to detect such objects: https://github.com/rozumden/FMODetect