Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5

Teaching AI to see depth in photographs & paintings + Wildfire simulation heats up

#1
C C Offline
Wildfire simulation heats up
https://discovery.kaust.edu.sa/en/articl...n-heats-up

RELEASE: Computer simulations can predict the spread of wildfires through sections of a real forest using a realistic yet computationally efficient new method to model the combustion of individual trees. The model, which accurately captures wildfire behavior at forest scale, was created by KAUST researchers and their international collaborators.

“Our work can help to make real forest fires more predictable by simulating potential fire scenarios with real forestry data,” says Torsten Hädrich, a Ph.D. student in Dominik Michels’ group. “We are also able to simulate how firebreaks in forests can be used to contain fire spread,” he adds.

Climate change simulations have long predicted a global rise in the prevalence and intensity of wildfires. “We titled our research paper ‘Fire in Paradise’ in reference to the northern Californian town of Paradise that was devastated by a wildfire in November 2018, resulting in more than 80 fatalities,” says Michels, who led the research.

Accurate computational simulations could improve our understanding and management of wildfire behavior. “The main challenge is to capture the complex dynamics involved,” Hädrich explains. The simulation must include a model for tree combustion and a fluid dynamics component for the simulation of fire while accounting for environmental variations, such as tree density, terrain and wind.

Previous forest-scale fire simulations have represented trees as simple cones or cylinders, but this extreme simplification of forest structure reduces the accuracy of the simulation. Hädrich, Michels and their collaborators developed a method to model trees as collections of branch-like modules. “Our module-based representation provides a way to control the level of detail for the simulation,” Hädrich says.

Depending on the scenario being simulated, each tree can be represented either by many detailed modules to generate highly realistic branching structures or by fewer coarser modules for efficient computation. “Our approach is capable of realistically simulating the propagation of fire through entire ecosystems with varying forest cover,” Hädrich says.

Further improvements of the predictive accuracy of the simulation include modeling the cloud formation and rainfall that can result as water vapor is released from burning vegetation. Additional refinements to the simulation are planned, notes Michels. “We would like to include the simulation of sparks flying through the air and address fire spread on the ground facilitated by grass, branch litter and undergrowth vegetation,” he says. “We also plan to further validate our approach using satellite images of real wildfires,” Michels adds.



Teaching AI to see depth in photographs and paintings
https://educationnewscanada.com/article/...tings.html

RELEASE: For humans, determining which objects are close and far away when we look at a photograph is effortless. While we take this ability for granted, it's quite challenging for computers. This may not be the case for long, as researchers in the Computational Photography Lab at SFU are successfully teaching artificial intelligence (AI) how to determine depth from a single photograph, as discussed in their recent CVPR 2021 publication.

This process, called monocular depth estimation, has many applications in image editing and beyond. "When we look at a picture, we can tell the relative distance of objects by looking at their size, position, and relation to each other," says Mahdi Miangoleh, an MSc student in the Computational Photography Lab. "This requires recognizing the objects in a scene and knowing what size the objects are in real life. This task alone is an active research topic for neural networks."

In recent years, there has been a lot of progress on this research topic, but existing works fail to give a high-resolution input that is ready to transform the image into a 3-dimensional (3D) space. To counter this, the lab's approach is about utilizing the untapped potential of existing neural network models.

The proposed research explains the lack of high-resolution results in current methods through the limitations of convolutional neural networks. Despite major advancements in recent years, the neural networks still have a relatively small capacity to generate many details at once. Another limitation is the amount of pixels these networks can look at' at once, which determines how much information the neural network can make use of to understand complex scenes.

"We analyze the image and optimize our process by looking at the image content according to the limitations of current architectures," says Sebastian Dille, a PhD student in the same lab. "We give our input image to the network in many different forms, to create as many details as the model allows while preserving a realistic geometry."

By increasing the resolution of the estimation, the researchers make it possible to create detailed 3D renderings that look realistic to a human eye. Depth maps are used to create 3D renderings of scenes and simulate camera motion in computer graphics. "With the high-resolution depth maps that we are able to get for real-world photographs, artists and content creators can now immediately transfer their photograph or artwork into a rich 3D world," says computing science professor Yağız Aksoy, who leads the Computational Photography Lab.

Aksoy points to an interesting application of being able to estimate depth from a single image the same method can also generate realistic depth maps of paintings! This is not possible with traditional geometry estimation methods in computer vision which require multiple cameras looking at the same object. Since a painting is a 2D object, only single-image depth estimation can see' depth within it.

We are already seeing artists around the world utilize the applications enabled by Aksoy's lab's research. Akira Saito, a visual artist based in Japan, is creating videos that take viewers into fantastic 3D worlds dreamed up in 2D artwork. To do this he combines tools such as Houdini, a computer animation software, with the depth map generated by Aksoy and his team.

Creative content creators on TikTok are already using similar research to express themselves in new ways. This tutorial explains the process of how high-resolution depth maps can be used to create impressive 3D visualizations of photographs. It also shows how this research can be used to create interesting snippets.

"It's a great pleasure to see independent artists make use of our technology in their own way," says Aksoy. "We have made great leaps in computer vision and computer graphics in recent years, but the adoption of these new AI technologies by the artist community needs to be an organic process, and that takes time."

The next steps in this lab's research are extending this work to videos and developing new tools that will make depth maps more useful for artists. (RELATED)

PAPER: https://openaccess.thecvf.com/content/CV...paper.html

Boosting Monocular Depth Estimation to High Resolution https://youtu.be/lDeI17pHlqo

https://www.youtube-nocookie.com/embed/lDeI17pHlqo
Reply


Possibly Related Threads…
Thread Author Replies Views Last Post
  How to test if we’re living in a computer simulation C C 11 377 Nov 27, 2022 07:53 AM
Last Post: Kornee
  Coevolution of particle physics & computing + Female pioneer of computer simulation C C 0 76 Oct 1, 2021 03:43 PM
Last Post: C C
  Rise of AI art - what it means + Teaching AI to connect senses like vision & touch C C 0 197 Jun 17, 2019 04:34 PM
Last Post: C C
  Universe Is Not a Simulation, but We Can Now Simulate It C C 0 250 Jun 18, 2018 03:08 AM
Last Post: C C
  Researchers claim we are NOT living in a simulation C C 1 493 Oct 3, 2017 05:26 PM
Last Post: Zinjanthropos
  Round 2 - 'Elon Musk Is Wrong: We Aren't Living in a Simulation' C C 2 836 Jun 30, 2016 04:53 AM
Last Post: C C
  Improved simulation of neurons elte 1 662 Mar 6, 2016 06:44 PM
Last Post: C C
  How researchers are teaching computers the laws of physics C C 0 931 Jan 7, 2016 10:41 AM
Last Post: C C
  The Argument for Teaching Computer Science Without Computers C C 0 622 Jun 4, 2015 02:48 PM
Last Post: C C



Users browsing this thread: 1 Guest(s)