#cv

22 Oct 2026 · Paper summary
Paper summary: Spectral-Spatial Interpretation
Spectral-Spatial Interpretation explains multispectral image models with a full position × channel attribution matrix, using a local Shapley game at each position, so it shows which channels matter where.
10 Jun 2026 · Paper summary
Paper summary: ABot-Earth 0.5: Generative 3D Earth Model
ABot-Earth 0.5 generates large, seamless 3D Gaussian Splatting environments from satellite images alone, at under 10 minutes per square kilometre, viewable in web map engines and usable as a simulator for UAV navigation.
4 Mar 2026 · Paper summary
Paper summary: Utonia: Toward One Encoder for All Point Clouds
Utonia trains one self-supervised point-transformer encoder across very different point-cloud domains (remote sensing, LiDAR, indoor RGB-D, CAD, video-lifted points), and its features also help robot policies and spatial reasoning in VLMs.
13 Jan 2026 · Paper summary
Paper summary: BabyVision: Visual Reasoning Beyond Language
BabyVision tests core visual abilities that don't depend on language, which young children handle easily; the best model tested (Gemini3-Pro-Preview) scores 49.7, behind 6-year-olds and far behind adults' 94.1.
18 Aug 2025 · Paper summary
Paper summary: DINOv3
DINOv3 scales self-supervised vision training and introduces Gram anchoring to stop dense feature maps from degrading during long training, producing a frozen backbone that beats specialised models across many tasks.
22 Oct 2024 · Paper summary
Paper summary: FrugalNeRF: Fast Convergence for Extreme Few-shot Novel View Synthesis without Learned Priors
FrugalNeRF trains NeRFs from very few views quickly, using weight-sharing multi-scale voxels and pseudo depth chosen by cross-scale reprojection error, without relying on learned priors.
2 Aug 2024 · Paper summary
Paper summary: SAM 2: Segment Anything in Images and Videos
SAM 2 extends Segment Anything to video with a streaming-memory Transformer and a data engine that built the largest video segmentation dataset to date.
10 Jul 2024 · Paper summary
Paper summary: Vision language models are blind: Failing to translate detailed visual features into words
BlindTest's seven simple visual tasks, such as whether two circles overlap or how many times two lines cross, trip up strong vision-language models: four state-of-the-art VLMs averaged 58.07%.
14 Jun 2024 · Paper summary
Paper summary: Depth Anything V2
Depth Anything V2 trains a monocular depth teacher on synthetic images only, then teaches student models through large-scale pseudo-labelled real images, giving finer and more robust depth than V1.
15 Mar 2024 · Paper summary
Paper summary: MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training
Apple's MM1 ablates how to build multimodal LLMs and finds the data mix and image encoder (resolution, token count) matter a lot, while the vision–language connector design matters little.
3 Dec 2023 · Paper summary
Paper summary: Merlin:Empowering Multimodal LLMs with Foresight Minds
Merlin trains a multimodal LLM to predict object trajectories and then reason about future events, using two stages called Foresight Pre-Training and Foresight Instruction-Tuning.
13 Nov 2023 · Paper summary
Paper summary: Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks
Florence-2 is a single sequence-to-sequence vision model that takes a text prompt as the task instruction and answers in text, covering captioning, detection, grounding and segmentation, trained on 5.4B annotations.
29 Sept 2023 · Paper summary
Paper summary: Vision Transformers Need Registers
ViTs repurpose some background patch tokens as internal scratch space, which produces high-norm artifacts; adding a few extra "register" tokens fixes this and smooths attention maps.
9 Aug 2023 · Paper summary
Paper summary: 3D Gaussian Splatting for Real-Time Radiance Field Rendering
3D Gaussian Splatting represents a scene as a large set of optimised 3D Gaussians and renders them with a fast rasteriser, reaching real-time novel-view synthesis at 1080p.