#VisionLanguageModel
LightNews LightNews
michabbb.bsky.social
Sub-1B #Vision Language Model: Introducing OmniVision-968M 🔍

#NexaAI introduces #OmniVision, 968M #VisionLanguageModel for edge devices with 9x token reduction & enhanced accuracy via #DPO. Based on #Qwen & #SigLIP architecture. Try demo on #HuggingFace

nexa.ai/blogs/omni-v...

#ai
OmniVision-968M: World's Smallest Vision Language Model
Pocket-size multimodal model with 9x token reduction for on-device deployment
nexa.ai
November 26, 2024 at 8:22 AM
ndif-team.bsky.social
NNsight 0.6 also introduces first-class support for VisionLanguageModel (e.g., LLaVA, Qwen-VL) and DiffusionModel (e.g., Stable Diffusion, Flux)! Available remote on NDIF soon 👀
February 27, 2026 at 5:14 PM
thedailytechfeed.com
Satellite uses AI to autonomously detect objects in orbit, revolutionizing space monitoring. #SatelliteAI #VisionLanguageModel #SpaceTech #EarthObservation #AIinSpace #TechInnovation thedailytechfeed.com/satellite-ac...
June 15, 2026 at 1:00 PM
mpbontenbal.eurosky.social
Can you ask questions about an image?

A 'Vision Language Model' links your question with the visual content of an image. It will generate a full response to your question.

#learnAI #VisionLanguageModel
April 17, 2025 at 11:38 AM
tsumari-i.bsky.social
判断に困っていて、いまだに「生成AI」とだけ呼ぶ人もいるため私もそのままでした。

「LLM」呼びする人が増えたきっかけになったnote記事も見ましたが、指摘をしながらも自身はそのLLMを使っている生成AIユーザーなことなども気になってこの呼び変えをするべきか悩んでます。

画像のVLM（VisionLanguageModel 視覚言語モデル）や音声のALM（Audio Language Mode 音声言語モデル）も含めたものを言語のLLM（LargeLanguageModel 大規模言語モデル）に含めて呼ぶべきなのか、どう呼ぶのが妥当なのか、調べても人によっても違うような？
June 28, 2026 at 2:21 AM
techchilli.bsky.social
Google DeepMind’s PaliGemma: A Small But Mighty Open-Source Vision-Language Model.

See here - techchilli.com/news/google-...

#GoogleDeepMind #PaliGemma #VisionLanguageModel #AI #TechInnovation #OpenSource #MachineLearning #AIEfficiency #TechTrends #FutureOfAI #ArtificialIntelligence
July 14, 2024 at 4:07 PM
michabbb.bsky.social
#MistralAI Document #AI: Advanced #OCR solution for complex document processing 📄

📺 www.youtube.com/watch?v=yrx...

🔧 Fine-tuned #VisionLanguageModel specifically designed for document understanding beyond traditional #OCR limitations that plague most business workflows

🧵 👇
July 22, 2025 at 8:15 PM
ronan.mastodon.ronandev.ovh.ap.brid.gy
Pour essayer Qwen2.5-VL :
1. Installer Ollama https://ollama.com/download
2. Télécharger/lancer le modèle : ollama run qwen2.5vl:7b
3. Exemple de prompt : Describe this picture /path/to/file.png

#opensource #vlm #llm #visionlanguagemodel
Download Ollama on macOS
Download Ollama for macOS
ollama.com
May 26, 2025 at 7:28 AM
voqn.bsky.social
記事中にもあるような「画像は、作られたものが1人歩きしてしまう側面があると思います。人間が作ったものなのか、AIが作ったものなのか、よほど注釈が入っていないと分からないので、有害となる可能性があります」の、「画像生成AIはやらない」は尊重する。

ただ、テキスト生成AIであれOCRとVisionLanguageModelを持つ時点で「有害なものを送りつけてくる」ケースに備えて、結局は"なんでも"潜在空間中に取り込ませなきゃいけないだろう。
platform.claude.com/docs/en/buil...

実際、制限には人物特定に関する指示や露骨な画像を受け付けないようになってる
Vision
Claude's vision capabilities allow it to understand and analyze images, opening up exciting possibilities for multimodal interaction.
platform.claude.com
January 7, 2026 at 1:27 PM
rehaistics-lab.bsky.social
電子健康記録（EHR）を活用した眼底疾患診断のためのVision-Language Pre-trainingモデル「Medrecord-CLIP」に関する研究。従来の視覚のみのモデルは固定的なカテゴリラベルに依存し、EHR内の個別患者情報を活かせていなかった。本手法はEHRを活用した医療ナラティブをテキスト監督に組み込むことで、網膜疾患スクリーニングの精度向上と臨床的文脈の反映を実現する。

#医療AI #眼科AI #VisionLanguageModel

https://pubmed.ncbi.nlm.nih.gov/42632973/
August 25, 2026 at 1:05 AM
narrowit.bsky.social
IREX Launches Beta Tool for Prompt-Based Video Detection IREX introduced StreamVLM, a beta vision-language system that turns text prompts into...

https://www.narrowit.com/news/irex-launches-beta-streamvlm-prompt-video-detection-2026-09-13
#IREX #Streamvlm #VisionLanguageModel #VideoAnalytics
IREX Launches StreamVLM Prompt-Based Video Detection Beta | Narrowit.com
IREX says StreamVLM can create video detectors from plain-language prompts for selected public-safety installations, with local GPU processing.
www.narrowit.com
September 13, 2026 at 4:20 PM
aidailypost.com
🚀 Cohere’s new Parse 5 just crushed ParseBench with a 79.2 score—beating GPT‑5.5 and Gemini 3.5 Flash on cost while nailing OCR to structured Markdown. Curious how this vision‑language model reshapes document parsing? Dive in! #Parse5 #ParseBench #VisionLanguageModel

🔗
August 28, 2026 at 3:45 PM
envdatascience.bsky.social
New article!

SYNOPTICBENCH: evaluating vision-language models on generating weather forecast discussions of the future

👉https://cup.org/4x01Dkh
✍️Timothy Higgins, Antonios Mamalakis and Chirag Agarwal

#visionlanguagemodel #multimodality #weatherforecasting
July 24, 2026 at 10:01 AM
getnews-me.bsky.social
HiViS cuts the drafter’s prefill sequence to just 0.7%‑1.3% of the original input, delivering up to 2.65× faster inference without quality loss in multimodal AI. Read more: https://getnews.me/hivis-boosts-vision-language-model-speed-with-visual-token-hiding/ #hivis #visionlanguagemodel
HiViS Boosts Vision-Language Model Speed with Visual Token Hiding
September 30, 2025 at 3:59 PM
hackernoon.com
Explore a collection of visualizations demonstrating the effectiveness of promptable and open-vocabulary segmentation across various datasets. #visionlanguagemodel
Visualizing Promptable and Open-Vocabulary Segmentation Across Multiple Datasets
hackernoon.com
November 13, 2024 at 6:00 PM
hackernoon.com
This evaluation explores promptable segmentation using uniform point grids and ground-truth bounding boxes across various datasets. #visionlanguagemodel
Evaluating Promptable Segmentation with Uniform Point Grids and Bounding Boxes on Diverse Datasets
hackernoon.com
November 13, 2024 at 5:00 PM
hackernoon.com
Uni-OVSeg combines CLIP, multi-scale pixel decoders, and visual prompts for effective open-vocabulary segmentation, boosting weakly-supervised learning. #visionlanguagemodel
Advanced Open-Vocabulary Segmentation with Uni-OVSeg
hackernoon.com
November 13, 2024 at 4:01 PM
hackernoon.com
Uni-OVSeg advances open-vocabulary segmentation, benefiting sectors like medical imaging and autonomous vehicles while addressing the risk of AI bias in dataset #visionlanguagemodel
Uni-OVSeg: A Step Towards Efficient and Bias-Resilient Vision Systems
hackernoon.com
November 12, 2024 at 10:27 PM
hackernoon.com
Uni-OVSeg outperforms weakly-supervised and fully-supervised methods in open-vocabulary segmentation, showing superior results on datasets like PASCAL and COCO. #visionlanguagemodel
Uni-OVSeg Outperforms Weakly-Supervised and Fully-Supervised Methods in Open-Vocabulary Segmentation
hackernoon.com
November 12, 2024 at 10:27 PM
hackernoon.com
Uni-OVSeg offers a breakthrough in open-vocabulary segmentation, reducing reliance on triplets and achieving superior performance, surpassing current models.
#visionlanguagemodel
Uni-OVSeg: Weakly-Supervised Open-Vocabulary Segmentation with Cutting-Edge Performance
hackernoon.com
November 12, 2024 at 10:27 PM
hackernoon.com
The baseline for open-vocabulary segmentation uses image-text and image-mask pairs with the CLIP model for feature extraction. #visionlanguagemodel
he Baseline and Uni-OVSeg Framework for Open-Vocabulary Segmentation
hackernoon.com
November 12, 2024 at 10:27 PM
hackernoon.com
Explore the datasets, including SA-1B and image-text pairs, used for training open-vocabulary segmentation. #visionlanguagemodel
Datasets and Evaluation Methods for Open-Vocabulary Segmentation Tasks
hackernoon.com
November 12, 2024 at 10:27 PM
hackernoon.com
Explore the evolution of segmentation techniques, from semantic to open-vocabulary segmentation, and the role of vision-language models in improving performance #visionlanguagemodel
The Future of Segmentation: Low-Cost Annotation Meets High Performance
hackernoon.com
November 12, 2024 at 10:27 PM
hackernoon.com
Get familiar with the open-vocabulary segmentation problem, where the aim is to segment images into masks associated with unseen semantic categories. #visionlanguagemodel
Defining Open-Vocabulary Segmentation: Problem Setup, Baseline, and the Uni-OVSeg Framework
hackernoon.com
November 12, 2024 at 10:27 PM
michabbb.bsky.social
#UITARS Desktop: The Future of Computer Control through Natural Language 🖥️

🎯 #ByteDance introduces GUI agent powered by #VisionLanguageModel for intuitive computer control

Code: lnkd.in/eNKasq56
Paper: lnkd.in/eN5UPQ6V
Models: lnkd.in/eVRAwA-9

#ai

🧵 ↓
January 22, 2025 at 6:34 PM