How to create Matterport3D segmentation images?
· ☕ 1 min read
Intro The other day, one of my labmates needed to make a segmentation of Matterport3D. He asked for help, and I got involved in creating the segmentation. However, it turned out to be a real struggle. We were not used to 3D mesh models. After several weeks, we completed the code to create a semantic segmentation image for Matterport3D. How to create Matterport3D segmentation images Matterport3D provides access to 3D segmentation but does not give users an easy way to access 2D. Matterport3D data only provides point clouds and meshes labeled by ground truth, and the user must add color directly to the point clouds and meshes to create 2D segmentations. We, therefore, wrote code using Matterport3DSimulator to place a camera for a given scan_id and viewpoint_id and create a segmentation from the original ply file. When we run our code, we get the following image. (I concatenated the obtained images and converted to a gif) Matterport3DSimulator takes a total of 36 pictures: 12 at the top, 12 at the perimeter, and 12 at the bottom. ...

· ☕ 1 min read
Matterport3DSimulatorをCUDA11.1で動かすDockerfile 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 rm /etc/apt/sources.list.d/cuda.listRUN rm /etc/apt/sources.list.d/nvidia-ml.listRUN apt-key del 7fa2af80RUN apt-key adv --fetch-keys apt-key adv --fetch-keys apt-get updateRUN apt-get -y upgradeRUN apt-get -y install nano wget curl# ONNX Runtime Training Module for PyTorch# Copyright (c) Microsoft Corporation. All rights reserved.# Licensed under the MIT License.ARG TORCH_CUDA_VERSION=cu111 ARG TORCH_VERSION=1.8.1ARG TORCHVISION_VERSION=0.9.1# Install and update tools to minimize security vulnerabilitiesRUN apt-get updateRUN apt-get install -y software-properties-common wget apt-utils patchelf git libprotobuf-dev protobuf-compiler cmake RUN unattended-upgradeRUN ...