Daria Romanska’s Post

Have you ever wondered how the dragons in Game of Thrones are brought to life? Or how a Sci-Fi Space station scene is created? In our latest video, we explore the rendering algorithm inside Computer Generated Images [CGI] and how the ray tracing algorithm creates realistic lighting and brings 3D scenes to life.

📷 Exciting developments in my computer vision journey!I've been diving deep into reverse image formation. With the camera matrix, I've successfully mapped 2D images back to their real-world coordinates. Now, I'm taking it a step further—predicting projectile paths in 3D space! 🚀 This is still a fresh approach for me, but the possibilities are endless. Imagine accurately predicting the trajectory of objects in real-time using #ComputerVision and #YOLO. Stay tuned for more updates as I refine this technique! https://lnkd.in/g4pTp9Cf

🌍 Best Tools for Visualizing Historical Data! Discover the best tools that allow researchers to explore historical maps, track boundary changes, and access rich contextual information. 1. GeaCron. Track historical borders with an interactive timeline. 2. Running Reality. Observe real-time boundary changes through dynamic animations. 3. Old Maps Online. Search and overlay historical maps on modern ones. Discover more in the full article https://lnkd.in/edyexTBW #historytools #aitools #historicaldata

GTA: Prague in real Prague? Mosaic - Geospatial Imaging Leaders has made freely available a huge dataset capturing most of Prague's streets. It consists of a total of 1,260,000 (!) panoramic photos. So theoretically, anyone can now create a model of Prague and walk (or drive) through it. The level of detail is comparable with Google Street View but unlike Street View, you can use it freely (non-commercially). For example, using the so-called Gaussian splatting, you can create a 3D Prague for virtual reality, virtual production, imagination is the limit. 🤯 Exciting times! Video by Mosaic. via Bilawal Sidhu

Recently I stumbled into the world of Gaussian Splatter and I am amazed by the possibilities that the future holds for 3D imagery. Here is a Vineyard that I had previously captured for mapping and I decided to try it out with. -Aligned with Reality Capture -Processed with Postshot. You can easily see all the damage caused by the winter frost. https://lnkd.in/gC9grWt8

First lesson in photogrammetry. Patience. After all of the hours of tweaking your alignment, processing your images, double checking settings. All you can do is wait...

HGS1 animated video of the Historic Commercial Double Lunar Flyby Rescue Mission on its 26th anniversary. Shout Outs to: Hughes Global Services, HGS, management for their courage to sponsor the mission; Analytical Graphics Inc., AGI, whose Satellite Toolkit, STK, was essential for planning the mission and whose graphical expertise allowed the world to follow the mission in real time on the internet; Van Martin Systems Inc., whose Microcosm software assimilated all types of orbital data from contributors around the world to tell us where we were at any time; and to the dozens of in house Hughes volunteers and contract personnel from all subsystems who flawlessly executed the 60 maneuvers required over the 69 day saga. The 6.5 minute animated video can be viewed in 3 segments for your choice 1) The 1st lunar flyby maneuvers/trajectory (~2,5 min) 2) Visual of the 2nd lunar flyby and the 2nd lunar flyby trajectory (~1.5 min) 3) Trajectory of the complete double lunar flyby (2 viewing angles,~2.5 min) 💌

Maxar's satellite imagery has been used as a map reference for vision based navigation on aerial platforms for years and remains one of the effective methods for GPS denied navigation. We believe we can supercharge vision based navigation with Maxar's Precision3D product. This capability will allow aerial systems to fly lower, faster, and over more complex environments. This edge capability is particularly important in contested environments where the EW spectrum is jamming not only GPS but also radio comms, significantly limiting traditional navigational methods.

How can we generate high-fidelity, complex 3D scenes? Quan Meng's LT3SD decomposes 3D scenes into latent tree representations, with diffusion on the latent trees enabling seamless infinite 3D scene synthesis! Check it out: https://lnkd.in/eM4_Y_yD Great work by Quan Meng, with Lei Li and Matthias Niessner

Some of my favorite analyses that I showcased at #NZEUC this year 1️⃣ STAC Connection + LINZ S3 Bucket - Historical Image Object detection: We tapped into both historical and latest imagery, connecting directly to LINZ’s S3 bucket. The impressive part? Using a single model (TextSAM) to process both RGB and black & white images, even with vastly different resolutions (75cm and 10cm). 2️⃣ NZ Land Cover Classification for Impervious Areas: Using Drone Imagery from our customer #Recon and #ArcGISRealityforArcGISPro, we created a #TrueOrtho, which was then stored on our ImageServer. With Deep Learning Studio, we can run inferencing and share the result back to the desktop #ArcGISPro for in-depth analysis to identify impervious areas. Although our model was trained with ortho mosaic, it performed exceptionally well with true ortho too! 💡 BONUS: Impervious Area Analysis for Matamata, NZ! Matamata not just famous for its proximity to the iconic Hobbiton Movie Set from Peter Jackson’s #LordoftheRings and #TheHobbit films, but now also for our latest analysis. #GeoAI #ArcGIS #STAC