MediaPipe
PulseAugur coverage of MediaPipe — every cluster mentioning MediaPipe across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
Google AI Classroom uses Gemini 4 Argon and Gemma 4 E2B for multimodal orchestration
This article details a multimodal orchestration strategy for AI-powered classrooms, utilizing a range of Google's Gemini models and Gemma. The approach emphasizes right-sizing models based on specific educational tasks,…
-
AI research tackles low-resource sign languages with new datasets and methods · 2 sources tracked
Two recent research papers address the significant challenge of developing AI for low-resource sign languages. The first paper, though withdrawn, proposed pathways for languages like Azerbaijan Sign Language, emphasizin…
-
New method generates blendshapes for 3D models with disconnected parts
Researchers have developed a novel method for generating blendshapes for 3D character models, specifically addressing the challenge of models with disconnected parts. Their approach utilizes neural networks to create de…
-
On-device LLMs gain traction for Flutter apps in 2026
The trend in 2026 is shifting towards on-device AI for mobile applications, moving away from cloud-based APIs like OpenAI and Anthropic. This approach offers enhanced privacy, eliminates API costs, and enables offline f…
-
Fine-tuning Gemma 2 for On-Device LLM Tasks
This article details the process of fine-tuning the Gemma 2 model using 4-bit LoRA for efficient on-device execution. The author shares their experience running these smaller LLM tasks offline on a phone with MediaPipe,…
-
Phone-side workout coach developed with MediaPipe and generative AI
A phone-side workout coach was developed using MediaPipe and generative AI technologies at Innovhacks 4.0. This project was shared on the DEV Community platform, highlighting its use of software development and coding t…
-
emg2face uses EMG signals for expressive facial animation in VR
Researchers have developed a novel method called emg2face that uses high-density surface electromyography (HD-sEMG) to generate expressive facial animations, bypassing the need for optical face capture. This technique i…
-
Simon Willison builds privacy tool with GPT-6 Astra
Simon Willison developed an experimental tool called Photo Scrubber to automatically blur faces and remove metadata from photographs. The tool utilizes Google's MediaPipe library compiled to WebAssembly and the BlazeFac…
-
Low-cost reservoir computing model advances sign language recognition
Researchers have developed a novel, low-cost hybrid reservoir computing model for recognizing isolated sign language videos. This model utilizes MediaPipe to extract key body and hand points, which are then processed by…
-
New RSC-GestureNet system enhances traffic gesture recognition for autonomous driving
Researchers have developed RSC-GestureNet, a new system designed to reliably recognize Chinese traffic police gestures for autonomous driving applications. This model incorporates pose confidence as a key factor, down-w…
-
OpenAI's GPT Image 2 simplifies image editing, replacing complex tools
OpenAI's GPT Image 2, also known as ChatGPT Images 2.0, offers a streamlined approach to image editing, replacing the need for complex local setups like ADetailer with Stable Diffusion. Users can now make single-frame e…
-
Teenager builds fully local AI ecosystem CODA OS
A 17-year-old developer has created CODA OS, a comprehensive AI ecosystem designed to run entirely on local hardware without cloud dependencies. The system includes a 3D reconstruction tool, a custom 2.2B parameter lang…
-
New Transformer Model Enhances Face Recognition with Masked Faces
Researchers have developed PLGSA-Transformer, a novel framework for face recognition that addresses the challenges posed by facial masks. This system utilizes periocular landmark-guided spatial attention to focus on vis…
-
Google's webcam hand-scan reCAPTCHA quickly bypassed by testers · 3 sources tracked
Google is testing a new reCAPTCHA system that uses a webcam to scan users' hands, mapping 21 points to verify human identity. This experimental feature, part of Google Cloud Fraud Defense, aims to combat bots more effec…
-
New Gated Affect Transformer improves human motion prediction
Researchers have developed a new method called the Gated Affect Transformer (GAT) to improve human motion prediction by integrating facial affect cues with body pose data. The study found that simply combining these mod…
-
AI touch detection framework struggles with real-world mobile typing reconstruction
A new research paper details a multi-modal framework for detecting touch events on mobile keypads using video surveillance. The system integrates hand landmark detection, skin color filtering, motion detection, and edge…
-
Virtual ring try-on system uses AI for realistic placement
Researchers have developed a novel virtual ring try-on system that allows users to see how a ring would look on their hand. The system uses MediaPipe for hand point detection and YOLO-V8 for ring object detection to acc…
-
New AI tool Envisage visualizes rhinoplasty outcomes with novel evaluation metric
Researchers have developed Envisage, a new pipeline for visualizing the intended outcomes of rhinoplasty surgery using diffusion-based generative editing. This system is designed to provide localized edits and includes …
-
Withdrawn research paper details CPU-based fall detection system
A research paper, now withdrawn, proposed a fall detection system for elderly care that operates efficiently on standard CPUs. The system utilizes pose estimation via the MediaPipe framework to analyze motion and body p…
-
AI framework digitizes athlete profiling with VLM and RAG
Researchers have developed a novel LLM-based framework for holistic athlete profiling, designed to overcome the limitations of traditional manual or basic computer vision assessment methods. This agentic system, orchest…