#25
The behavior is same as the issue listed above.
The code seems to be getting stuck in the following function
def tracking(self, pipe):
self.tracker = Tracker(self, pipe)
self.printer.print("Tracking Triggered!", FontColor.TRACKER)
self.all_trigered += 1
os.makedirs(f"{self.save_dir}/mono_priors/depths", exist_ok=True)
os.makedirs(f"{self.save_dir}/mono_priors/features", exist_ok=True)
while self.all_trigered < self.num_running_thread:
pass
self.printer.pbar_ready()
self.tracker.run(self.stream)
self.printer.print("Tracking Done!", FontColor.TRACKER)
This function is in slam.py
I am running on A40 machine with 24 vCPU, Ubuntu 22.04, CUDA 11.8, 48GB GPU RAM
It seems the self.all_trigered is always less than self.num_running_thread after about 6 to 7 percent of tracking phase of demo.