Clarifai’s Post

#10 DOFA Multimodality is one of the keys to remote sensing foundation models. Most of the methods, however, focus on just one of them. DOFA employs an innovative approach utilizing wavelength as a unifying parameter across various EO modalities to achieve a more cohesive multimodal representation. DOFA is trained using a masked image modeling strategy, and a distillation loss is included to further optimize its performance. what I liked the most: the wavelength approach paper: https://lnkd.in/dTBSEHk4 #50papers #AI4EO #remotesensing #deeplearning #GeoAI

🚀 Excited to share our latest blog post on Gaussian-SLAM: Photo-realistic Dense SLAM with Gaussian Splatting. Our new dense SLAM method uses Gaussian splats as a scene representation, enabling interactive-time reconstruction and photo-realistic rendering of real-world and synthetic scenes. We propose novel strategies for seeding and optimizing Gaussian splats, extending their use to sequential monocular RGBD input data setups. Check out the full article here: https://bit.ly/41pSYJn. #SLAM #ComputerVision #GaussianSplatting

What happens during the process of augmenting physical objects? How does augmentation actually work? Before the camera can augment the image or the object, it has to recognize it first. Image and object recognition are broad terms that cover various computer vision tasks. Image recognition includes such components as image detection, pattern matching, content overlay, and tracking with interaction. Object recognition in its turn consists of object detection, feature extraction, matching & tracking, and content overlay with interaction. Slide into this week’s Knowledge Bites and check out our Knowledge Base for more information 👉 https://lnkd.in/d9NCMYik #SQUARS #WebAR #KnowledgeBites #ARmarketing #AR #education #technology

🔍 Unlock Geometry Secrets! 📐 Dive into today's question: In the figure, altitudes AD and CE of ∆ABC intersect at point P. Discover the fascinating relationships: (i) ∆AEP ~ ∆CDP (ii) ∆ABD ~ ∆CBE (iii) ∆AEP ~ ∆ADB (iv) ∆PDC ~ ∆BEC. Explore more at www.qmaths.com! #Geometry #QOTD #Mathematics

My capstone on object detection and image classification on stroke images.

We simply explain and illustrate Mamba and (Selective) State Space Models – SSMs. SSMs match performance of transformers, but are faster and more memory-efficient than them. This is crucial for long sequences! 📺 https://lnkd.in/en6Vapn9 Incredible work by Albert Gu and Tri Dao! 👏 #SSM #researchpaper #video #artificialintelligence

🔥 #hottopic Multiscale Pixel-Level and Superpixel-Level Method for Hyperspectral Image Classification: Adaptive Attention and Parallel Multi-Hop Graph Convolution by Junru Yin, et al. ➡️ https://brnw.ch/21wLoTt

A powerful tool for solving systems of equations, transforming geometric objects, and modeling real-world phenomena. #Mathematics #Matrix #Algebra #DataScience#TheClasses

"SEM-GAT: Explainable Semantic Pose Estimation using Learned Graph Attention" by Efimia Panagiotaki, Daniele De Martini, Georgi Pramatarov, Matt G., and Lars Kunze "This paper proposes a GNN-based method for exploiting semantics and local geometry to guide the identification of reliable pointcloud registration candidates. Semantic and morphological features of the environment serve as key reference points for registration, enabling accurate lidar-based pose estimation. Our novel lightweight static graph structure informs our attention-based keypoint node aggregation GNN network by identifying semantic instance-based relationships, acting as inductive bias to significantly reduce the computational burden of pointcloud registration. By connecting candidate nodes and exploiting cross-graph attention, we identify confidence scores for all potential registration correspondences, estimating the displacement between pointcloud scans. Our pipeline enables introspective analysis of the model's performance by correlating it with the individual contributions of local structures in the environment, providing valuable insights into the system's behavior. We test our method on the KITTI odometry dataset, achieving competitive accuracy compared to benchmark methods and a higher track smoothness while relying on significantly fewer network parameters." Paper: https://lnkd.in/dE2DXNHg #graphneuralnetworks #computervision

This video depicts the difference in Surface Reconstruction quality between Photogrammetry, Neural Radiance Fields (NeRFs), 3D Gaussian Splats (SuGAR) and the latest 2D Gaussian Splats. I have been testing the code for different approaches and thought it would provide a visual reference as to where all of these technologies stand as of this moment. The dataset contains 27 images. I will write a more detailed comparison between the pipelines in a later post.