Pedestrian360

The implementation of "Pedestrian-Aware Panoramic Video Stitching Based on a Structured Camera Aray".

Instructions

• calibration : calibration/calibrate_extrinsics: calibrate the extrinsics of the multi-camera system through the images in calibration/extrinsics_frames. The results are saved in calibration/results
• video_stitching: Run video_stitching/multiThread.cpp to stitch the videos.
• server: server/server.py receive the frames from the stitching program and perform human segmentation using Mask R-CNN. Then it will send the mask data to the stitching program.
• If you want to speed up the stitching, set the rescale factor in video_stitching/const.h and server/server.py at the same time.

Requirements

• Python 3.6
  • torch 1.3.0
  • torchvision 0.4.1
  • numpy
  • g2o
  • yaml
  • cv2
  • OpencGL
  • pangolin
• C++
  • cmake
  • OpenCV 4.2.0
  • eigen3
  • Threads