timzhang642 / 3d-machine-learning Goto Github PK

Hi,

Two papers on Scene Understanding:

• (ICCV'19) Holistic++ Scene Understanding: Single-view 3D Holistic Scene Parsing and Human Pose Estimation with Human-Object Interaction and Physical Commonsense, with project page in https://github.com/yixchen/holistic_scene_human
• (NIPS'19) PerspectiveNet: 3D Object Detection from a Single RGB Image via Perspective Points

Thanks.

To see a survey of RGBD datasets, check out Michael Firman's collection

collection link is broken

Hi,
Found this in CVPR 2018
PointGrid: A Deep Network for 3D Shape Understanding
http://openaccess.thecvf.com/content_cvpr_2018/papers/Le_PointGrid_A_Deep_CVPR_2018_paper.pdf
Code: https://github.com/47deg/pointgrid

Volumetric grid is widely used for 3D deep learning due to its regularity. However the use of relatively ower order local approximation functions such as piece-wise constant function (occupancy grid) or piece-wise linear function (distance field) to approximate 3D shape means that it needs a very high-resolution grid to represent finer geometry details, which could be memory and computationally nefficient. In this work, we propose the PointGrid, a 3D convolutional network that incorporates a constant number of points within each grid cell thus allowing the network to learn higher order local approximation functions that could better represent the local geometry shape details. With experiments on popular shape recognition benchmarks, PointGrid demonstrates state-of-the-art performance over existing deep learning methods on both classification and segmentation

A abstract helps to find out the papers that just suit the readers' taste.

https://github.com/cgtuebingen/Flex-Convolution

I'm a green hand on 3D point cloud recognition, and I'm finding some learning materials like this. But there are so many papers and I have a question about the principle that these papers are recorded.
Do these papers represent milestones and they are all ecellent compared to most other normal papers，or are these papers just collected in terms of their context without any standard ?
Because there are so much knowledge on 3D machine-learing that it consumes too much of my energy, I want to find what is more important for beginners, and I think many beginners also have this problem. It will be great for me if some papers are recommend in different levels，such as from one star to five stars.
Finally, I really appreciate your efforts.

Great list, thanks. The link & image in this are broken for me:

"SUNCG: A Large 3D Model Repository for Indoor Scenes (2017) [Link]"

Hi,
I am trying to do Segmentation task on 3d CAD Models.
However, I need to create the dataset for same. I would like to know if there is any tool to mark different segments on the CAD Model and generate the required files which can be used as dataset.
Any help is highly appreciated

Guys, with all the magnificent work done here is it possible to start bringing together the 3D ML and the traditional CAD programs (Creo, SolidWorks, Catia, etc...)
This would be a real breakthrough in machine design world, reinforced with ML capabilities

Although it's not strictly 3D (Contains only RGB images), and not all of the images are indoor, this dataset is very useful

Seems like the SUNCG Dataset has gone offline.

If there are any insight on why that is or if there are any other ways of obtaining the dataset, I would like to know.

How to classify and train 3d volume components using python without GPU machine ?
@ Arthurg27

Hi, there is a new work for automatic deformation transfer, found in Siggraph Asia 2018
Title: Automatic Unpaired Shape Deformation Transfer
Project: http://geometrylearning.com/ausdt/
Paper: http://www.geometrylearning.com/paper/Automatic2018.pdf
Code: https://github.com/gaolinorange/Automatic-Unpaired-Shape-Deformation-Transfer
It may be included in transfer section. Thank you!

Transferring deformation from a source shape to a target shape is a very useful technique in computer graphics. State-of-the-art deformation transfer methods require either point-wise correspondences between source and target shapes, or pairs of deformed source and target shapes with corresponding deformations. However, in most cases, such correspondences are not available and cannot be reliably established using an automatic algorithm. Therefore, substantial user effort is needed to label the correspondences or to obtain and specify such shape sets. In this work, we propose a novel approach to automatic deformation transfer between two unpaired shape sets without correspondences. 3D deformation is represented in a high-dimensional space. To obtain a more compact and effective representation, two convolutional variational autoencoders are learned to encode source and target shapes to their latent spaces. We exploit a Generative Adversarial Network (GAN) to map deformed source shapes to deformed target shapes, both in the latent spaces, which ensures the obtained shapes from the mapping are indistinguishable from the target shapes. This is still an under-constrained problem, so we further utilize a reverse mapping from target shapes to source shapes and incorporate cycle consistency loss, i.e. applying both mappings should reverse to the input shape. This VAE-Cycle GAN (VC-GAN) architecture is used to build a reliable mapping between shape spaces. Finally, a similarity constraint is employed to ensure the mapping is consistent with visual similarity, achieved by learning a similarity neural network that takes the embedding vectors from the source and target latent spaces and predicts the light field distance between the corresponding shapes. Experimental results show that our fully automatic method is able to obtain high-quality deformation transfer results with unpaired data sets, comparable or better than existing methods where strict correspondences are required.

It seems that the slack link is broken, it said that "invitation expired", "link was deactivated".

Hi,

It would really help if a curated list of tools for 3d CAD Models is also added in this list.

For Eg.

1. Tools for Segmenting the CAD Model
2. Tools for Converting the CAD Models from one format to another.

PS: I am still in the processing of finding tool for Segmenting a CAD Model to create a custom training dataset. Since this github repo was my first go to point, I was disappointed when i didn't fine any tools here. Hence this issue.

Hello, thank you for the good resources for 3D Computer Vision.

Could you add Mechanical Component Benchmark that is a large-scale mechanical component dataset?

These are the links to the project and paper for your reference.

Thank you.

i noticed this repo stops around 2019

what's the latest?

Hello, thank you for your sharing. When I was reading papers recently, I found that many papers use the NTU dataset. This dataset has 549 3D objects from 46 categories, but I did not find a download method for this dataset. Do you know the download link?

Hi,

thanks for putting those resources together.
perhaps you can add a link to this one?
https://github.com/3d-morphable-models/curated-list-of-awesome-3D-Morphable-Model-software-and-data

Best
Bernhard

I noticed you had The Space of Human Body Shapes: Reconstruction and Parameterization from Range Scans (2003)
but didn't have MPII Human Shape The open source himan shape sccpe tools and data set

The icons for "Multi-view Images", "Volumetric", etc are cute, but it is (1) impossible to use them for searching, and (2) hard to remember what they mean. Maybe use text keywords (tags) instead?

ECON: Explicit Clothed humans Optimized via Normal integration
It's one of the best models for 3d reconstruction of face and body🥇

您好！
请问有什么好的方法给.off格式的数据做分割和标签？

"This invite link is no longer active."
Could you please update the link?

Is this a science fiction ?

Hi Tim,

Really nice resource. I just wanted to point out there are some missing images (e.g. Topology-Varying 3D Shape Creation via Structural Blending (2014)).

-Ib

I was thinking maybe include other intresting resources like this paper UnrealCV. I am not sure where it fits. They also have a list of similar resources synthetic-computer-vision