Train 3D segmentation model using labeled 2D images and raw 3D data

Learning 3D Semantic Segmentation with only 2D Image Supervision
arXiv paper abstract
arXiv PDF paper

… there has been an explosion of raw 3D data collected from terrestrial platforms with lidar scanners and color cameras.

However, due to high labeling costs, ground-truth 3D semantic segmentation annotations are limited …

… investigate how to use only those labeled 2D image collections to supervise training 3D semantic segmentation models.

Our approach is to train a 3D model from pseudo-labels derived from 2D semantic image segmentations using multiview fusion.

… proposed network architecture, 2D3DNet, achieves significantly better performance (+6.2–11.4 mIoU) than baselines …

Stay up to date. Subscribe to my posts
Web site with my other posts by category


Photo by Kimon Maritz on Unsplash

Survey of deep learning to re-identify people seen by multiple cameras

Deep Learning Based Person Re-Identification Methods: A Survey and Outlook of Recent Works
arXiv paper abstract
arXiv PDF paper

… person re-identification (Re-ID) … goal is to retrieve persons with the same identity from different cameras.


State of AI Report 2021 (Benaich and Hogarth) (includes Computer Vision)

State of AI Report 2021

The State of AI Report analyses the most interesting developments in AI … Now in its fourth year, the State of AI Report 2021 is reviewed by AI practioners in industry and research, and…

Kaggle 2021 survey on ML and Data Science

State of Data Science and Machine Learning 2021

Kaggle Executive Summary of survey on Machine Learning and Data Science
Kaggle PDF's%20State%20of%20Machine%20Learning%20and%20Data%20Science%202021.pdf

* Half between 22 and 34 in age
* Live in India or U.S.
* Have a Master’s degree
* Learn…

Image classifier explains features used and how to modify them to change output

Explaining in Style: Training a GAN to explain a classifier in StyleSpace
arXiv paper abstract
arXiv PDF paper
Project page

Image classification models can depend on multiple … attributes … An explanation…

Improved multi-object tracking by associating all detection boxes

ByteTrack: Multi-Object Tracking by Associating Every Detection Box
arXiv paper abstract
arXiv PDF paper

Multi-object tracking (MOT) aims at estimating bounding boxes and identities of objects in videos.

Most methods … objects with low detection scores, e.g. occluded…

Make images in-focus at every pixel with the dual-pixel cameras in smartphones

Defocus Map Estimation and Deblurring from a Single Dual-Pixel Image
arXiv paper abstract
arXiv PDF paper

… a method that takes as input a single dual-pixel image, and simultaneously estimates the image’s defocus map — the…

Improve road lane detection by using multiple images with neural networks

A Hybrid Spatial-temporal Sequence-to-one Neural Network Model for Lane Detection
DeepAI paper abstract
arXiv PDF paper

Reliable and accurate lane detection is of vital importance for the safe performance of Lane Keeping Assistance and Lane Departure Warning…

Identify human actions on objects and their locations after training on image captions

Weakly Supervised Human-Object Interaction Detection in Video via Contrastive Spatiotemporal Regions
arXiv paper abstract
arXiv PDF paper

… introduce the task of weakly supervised learning for detecting human and object interactions in videos.

AI News Clips by Morris Lee: News to help your R&D

I apply innovative technologies like machine learning, computer vision, and physics to further an organization's goals. Am recognized innovator with 64 patents.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store