COCO
PulseAugur coverage of COCO — every cluster mentioning COCO across labs, papers, and developer communities, ranked by signal.
- used by Kingdom Hearts IV 90%
- used by Imagenet 1k 70%
- used by ADE20K 70%
- instance of Lvis 70%
- instance of Gotit.pub 70%
- instance of ScienceCast 70%
- instance of Imagenet 1k 70%
- used by Vision Transformers 70%
- instance of ADE20K 70%
- instance of CatalyzeX 70%
- used by Ms Coco 70%
- instance of alphaXiv 70%
14 day(s) with sentiment data
-
New S2-RL framework advances generative semantic segmentation dataset distillation
Researchers have introduced S$^2$-RL, a novel framework for generative semantic segmentation dataset distillation. This approach utilizes a diffusion model for text-to-image generation, allowing for the creation of nove…
-
New BUTR method optimizes tiny object detection with dynamic query budgeting
Researchers have developed a new method called BUTR (Budgeting-based DETR) to improve tiny object detection in images. This approach disentangles the effects of query selection and query quantity, focusing on dynamic qu…
-
New algorithm slashes AI optimization constraint violation to polylogarithmic levels
Researchers have developed a new algorithm for constrained online convex optimization that significantly reduces cumulative constraint violation (CCV) to polylogarithmic levels while maintaining near-optimal regret. Thi…
-
New metric evaluates multimodal AI data coherence beyond task accuracy
Researchers have introduced the Multimodal Coherence Score (MCS), a new metric designed to evaluate the quality of data fusion in multimodal AI systems. Unlike traditional methods that rely on downstream task accuracy, …
-
New framework uses open-vocabulary models for efficient, uncertainty-aware object detection annotation
Researchers have developed a semi-supervised framework to reduce the manual annotation effort required for object detection datasets. This method utilizes open-vocabulary models to generate object representations and th…
-
ERCLIP adapts CLIP for longer image-text retrieval with less data
Researchers have developed ERCLIP, a method to adapt CLIP for image-text retrieval with longer descriptions. ERCLIP fine-tunes CLIP by encoding long captions in segments and fusing these embeddings, improving retrieval …
-
Developer creates 'Coco' fan mod for Anthropic's Claude Code
A developer has created an unofficial fan modification called Coco, designed to enhance the user experience with Anthropic's Claude Code. Coco is a small, animated character that lives in the user interface, reacting to…
-
New method unlocks precise spatial control in image synthesis models
Researchers have identified a critical factor for improving spatial control in flow-matching models used for image synthesis: the signal-to-noise ratio (SNR) distribution during training. By shifting focus to high-noise…
-
New CS-CLIP Model Tackles "Half-Truths" in Image-Text Similarity
Researchers have identified a vulnerability in CLIP-style dual encoders, termed "half-truths," where appending plausible but incorrect details to a description can paradoxically increase similarity scores. This issue st…
-
LLMSeg pipeline uses LLMs for image segmentation without training
A new training-free pipeline called LLMSeg has been developed that leverages multimodal large language models (MLLMs) for image segmentation. This approach bypasses the need for dedicated segmentation architectures or t…
-
New method tackles relational OOD errors in vision-language models
Researchers have introduced a novel evaluation method called counterfactual pairing cycles to address the challenge of assessing relational out-of-distribution errors in vision-language models. This technique aims to pr…
-
Vision-Language Models Show High Confidence Despite Accuracy Drop in Degraded Visibility
A new study published on arXiv investigates the robustness of vision-language models (VLMs) by testing their performance under degraded visibility conditions. Researchers measured accuracy, confidence levels, and the mo…
-
Edge GPU AI performance bottlenecked by memory bandwidth, not FLOPs
A new research paper highlights that the performance of query-based segmentation decoders on edge GPUs is primarily limited by memory bandwidth rather than computational power (FLOPs). The study, which profiles models l…
-
New MTL-AL Framework Deconstructs Objectives, Improves Performance
Researchers have proposed a new framework for multi-task active learning (MTL-AL) that moves away from linear scalarization of objectives. The proposed method addresses the 'Conflict Paradox,' where penalizing gradient …
-
New research tackles LLM hallucinations with advanced detection methods · 5 sources tracked
Researchers are developing new methods to detect and mitigate hallucinations in large language models (LLMs). One approach, Cross-Layer Attention Probing (CLAP), processes LLM activations across its entire residual stre…
-
RT-DETRv4 framework boosts real-time object detection using Vision Foundation Models
Researchers have introduced RT-DETRv4, a new framework designed to enhance real-time object detection capabilities. This approach leverages Vision Foundation Models (VFMs) to improve lightweight object detectors without…
-
New CoCo framework enhances deep graph clustering with improved node representations
Researchers have introduced CoCo, a new framework designed to improve deep graph clustering by addressing limitations in how graph neural networks (GNNs) learn node representations. CoCo aims to capture both compactness…
-
New metric enhances interpretability of generative model evaluations
Researchers have introduced a new metric called directional Fréchet distance to better interpret the Fréchet distance, a common evaluation metric for generative models. This new metric decomposes the Fréchet distance in…
-
New distillation methods accelerate diffusion models, reducing computational cost · 3 sources tracked
Researchers have developed new methods to make diffusion models, which are powerful for generating content but computationally expensive, faster and more efficient. One approach, Sig-DEG, uses signature-based differenti…
-
FUSEye framework enhances fisheye camera object detection with minimal training
Researchers have developed FUSEye, a novel framework designed to improve the performance of object detection models on fisheye camera images. Fisheye cameras, commonly used in mobile robots, present challenges due to ra…