The preprocessing tool is dependent on the Python interface of FFmpeg ( ffmpeg-python), which in turn depends on the ffmpeg.exe and ffprob.exe files. If you need other versions, please get more information from the official FlowNet2-PyTorch website. PyTorch 1.3.1 for window_圆4, CUDA 9.2 (install instruction:.Pre-built *.pyd with following environment: The PyTorch python package contains all run-time dependencies of CUDA and CUDNN. The pre-built FlowNet2-PyTorch is included in the Python\preprocessing\flownet2 of this project. The pre-trained models can be downloaded from GitHub. The pre-processing script optionally uses flownet2-pytorch to compute optical flow.
Insta360tudio flowstate download#
COLMAPĭownload the pre-built COLMAP 3.6-dev.3 from GitHub, and download the Vocabulary Trees files Vocabulary tree with 1M visual words from the official COLMAP website. If you run OpenVSLAM on Windows you can download the modified OpenVSLAM's pre-built binaries from openvslam-bin-dd8af1c.
Insta360tudio flowstate install#
Pythonĭownload the installer from Python official website.Īfter setting up the python, run the following command to initialise the Python virtual environment and install the necessary packages of the preprocessing script.ĭownload the OpenVSLAM source code from the official website and set up following the official installation guide in Linux.
To keep the exported video upright, export the video with Insta360 Studio 3.4.2 with FlowState stabilisation and without Direction Lock. Get 'Insta360 Studio 2019' from the official website.Īt some point between versions 3.4.2 and 3.4.6, the FlowState stabilisation stopped keeping the camera up direction vertical, but instead tilts the video. Insta360 Studio 2019 Version: Windows v3.4.2.The default version of runtime 3rd-party software: The temlate config file python-config.yaml provided assumes that the OpenVSLAM and OmniPhotos binaries are in the same directory, so these should be copied to the dist/ directory.īelow are instructions to install and run the package using Python. Run this command and wait for completion:
Insta360tudio flowstate update#
Stitch the 360° video according to Section 2.1 below.Ĭopy the preproc-config-template.yaml template from the preprocessing binary folder to the same directory as the video, and update the preprocessing.input_path setting in it to the name of your video. FFmpeg: version 4.2.1 or newer (see Section 2.6 below)Īll other dependencies are provided in the package.This package contains all the necessary binaries and dependencies to generate a dataset from a stitched 360° video, and to view it. The red box is the preview window and the yellow is the logging window.Īn exported preprocessing binary for creating and viewing OmniPhotos can be downloaded from the Releases. The preprocessing application's GUI looks as follows.
This pre-processing pipeline prepares datasets for OmniPhotos preprocessing from videos captured with a 360° camera, such as an Insta360 One X.
0 kommentar(er)
