Grounding DINO
PulseAugur coverage of Grounding DINO — every cluster mentioning Grounding DINO across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New benchmark and method tackle shampoo bottle defect detection across SKUs
Researchers have introduced Bottle-SKU, a new benchmark dataset designed for detecting surface defects on shampoo bottles across various stock keeping units (SKUs). This dataset includes over 13,000 images, nearly 20,50…
-
Text-steerable visual encoders exhibit social bias, researchers find
Researchers have identified significant social biases in text-steerable visual encoders, specifically within the SteerViT model. These models, designed to point to image regions based on text prompts, were found to exhi…
-
New visual encoder SteerViT shows racial and gender bias, researchers find
Researchers have identified significant social biases in the SteerViT model, a text-steerable visual encoder. When prompted with demographic or crime-related terms, SteerViT disproportionately associates these with spec…
-
User seeks optimal AI-assisted annotation for Meta DETR fine-tuning
A user on Reddit is seeking advice on the most effective method for annotating approximately 500 images to fine-tune Meta's DETR object detection model. The primary challenge is balancing speed with accuracy, as manual …
-
GroundingPI model advances physical intelligence with visual primitives
A new 4-billion parameter grounding foundation model called GroundingPI has been introduced, aiming to advance physical intelligence through visual primitives. This model establishes a new state-of-the-art across 34 gro…
-
GroundAnything model achieves SOTA visual grounding with 4x speedup
Researchers have introduced GroundAnything, a new 4-billion parameter diffusion model designed for visual grounding tasks. This model utilizes blockwise denoising to achieve faster parallel decoding compared to traditio…
-
VLMs adapted for few-shot multispectral object detection
Researchers have adapted Vision-Language Models (VLMs) like Grounding DINO and YOLO-World for few-shot multispectral object detection. This approach integrates text, visual, and thermal modalities, demonstrating superio…
-
VTOS framework learns to orchestrate vision tools via adaptive search
Researchers have developed VTOS, a novel framework for orchestrating vision tools that adaptively searches for solutions and observers. This approach co-searches executable solution programs, which compose tools like Gr…
-
New ReVA model enhances visual question answering with region-aware AI
Researchers have developed ReVA, a novel region-aware visual assistant designed to improve multimodal large language models (MLLMs) in visually grounded question answering. ReVA addresses limitations in spatial reasonin…
-
WALDO system uses V-JEPA 2.1 features for one-shot object detection
Researchers have introduced WALDO, a novel one-shot object detection system designed for cluttered scenes. WALDO utilizes frozen features from the V-JEPA 2.1 world model, requiring only 3.4 million trainable parameters.…
-
New framework enhances privacy in urban traffic datasets
Researchers have developed a novel framework for automatically anonymizing urban traffic datasets, specifically addressing challenges in Kuala Lumpur, Malaysia. The system integrates Grounding DINO, a vision-language tr…
-
New frameworks enhance open-vocabulary object detection with domain adaptation and class geometry
Researchers are developing new frameworks for open-vocabulary object detection, which aims to improve model performance when encountering image domain shifts. PISA, a novel method, uses a Corruption-Invariant Feature Ex…
-
New GCRL method uses dynamic object masks for visual goal representation
Researchers have developed a novel approach for goal-conditioned reinforcement learning (GCRL) that utilizes dynamic object masks as goal representations. This method bypasses the need for privileged state or position i…
-
AI scientist defines 'world models' as action-conditioned, crucial for embodied AI
Zhang Lei, a leading AI scientist and founder of Visionary Future, argues that the current boom in "world models" within the AI industry is driven by the need to overcome the limitations of imitation learning in embodie…
-
New ABRA method transfers object detection knowledge across domains
Researchers have introduced ABRA (Aligned Basis Relocation for Adaptation), a novel method designed to transfer knowledge from labeled source domains to target domains lacking annotated data for open-vocabulary object d…
-
GeoAI tutorial details building footprint extraction using U-Net, DINO, SAM, and Mask R-CNN
This tutorial details a GeoAI workflow for extracting building footprints from aerial imagery using a combination of deep learning models. It covers setting up the geospatial environment, training a U-Net model with a R…
-
AI object detectors signal presence, not visibility, in cluttered scenes · 2 sources tracked
A new research paper reveals that open-vocabulary object detectors, widely used for tasks like grounding language and active perception, often signal the presence of an object rather than its visibility. Even when an ob…
-
User study finds improved robot interaction system perceptible to users
A new study published on arXiv explores the perceptual differences users experience when interacting with a multimodal human-robot system. The research compared a baseline system using Whisper, Florence-2, and Llama 3.1…
-
New AI framework translates radiologist speech to MRI tumor segmentation
Researchers have developed LoGSAM, a novel framework designed for parameter-efficient segmentation of brain tumors in MRI scans. This system transforms radiologist dictations into text prompts that guide foundation mode…
-
HKVLM model improves visual reasoning by separating localization from language
Researchers have developed HKVLM, a novel approach to visual reasoning that separates localization from language generation. This model utilizes a frozen language-aligned detector and a frozen language model, connected …