Ansys’ Post

With the rise of powerful third-party models, entry barriers have collapsed, but there's still a long journey from acceptable results to production-ready outcomes. If you're interested in Automatic Speech Recognition or Speech-to-Text, take a look at this 👇

🤖🛣️ Excited to share my latest blog post on guiding GenAI Agents with Semantic Router! Discover how leveraging a semantic routing layer can enhance reliability, accuracy, and speed in GenAI Agent workflows. Check it out here: https://lnkd.in/gsnRv-WY

Come along for the ride to learn about multimodality with Gemini! In this lab, you'll use multimodal prompts to extract info from text and visual data, generate a video description, and retrieve extra info beyond video using multimodality with Gemini → https://goo.gle/3W18clB

if you want to get a quick grasp of #Multimodality with #Gemini on #VertexAI do check out this lab from Google Cloud.

📣 Excited to introduce *locality alignment*, an efficient post-training algorithm that improves your ViTs + VLMs essentially for free! Local align = new self-supervised objective ensuring that encoder captures fine-grained spatial info. Paper: https://lnkd.in/g-EwjdJe Code: https://lnkd.in/gCFaBA87 Standard ViT is bad at "what's where". Locality alignment fine-tunes vision encoder to reconstruct the original ViT embeddings where image patches are randomly masked. This forces encoder to learn what's in each patch, greatly improving CLiP, SigLIP and downstream VLMs. Great work led by Ian Covert w/ collaborators Tony Sun and Tatsunori Hashimoto 👏

Just completed this eye-opening lab on Multimodality with Gemini on Google Cloud Skill Boost! This hands-on session explored the power of the Vertex AI Gemini API with Python using a Jupyter notebook on Vertex AI Workbench. We unlocked the potential of generating text descriptions, answering questions, and extracting information from images and even videos – all through creative prompts. Imagine using AI to: 👓 Recommend the perfect eyeglasses for your face shape ️ 🌏 Suggest your next travel destinations inspired by a captivating video ✨ Generate video descriptions that promote inclusiveness for all audiences Gemini makes these possibilities a reality, and I'm pumped about the future of generative AI and its disruptive potential across industries! What excites you most about generative AI? #VertexAI #GenerativeAI #MachineLearning P.S. If you don't have enough credits to start this lab, let me know and I can share some credits with you. 😎

This Google Cloud lab demonstrates a variety of multimodal use cases that #Gemini can be used for, including how to use the #VertexAI Gemini #API to generate text from text, images, and video prompts. #Gemini is a family of #GenAI models developed by Google Cloud #DeepMind designed for multimodal use cases. The Gemini #API provides access to Gemini Pro Vision and Gemini Pro models. intelia are a Google Cloud Premier Partner with a specialisations in Machine Learning and Data Analytics. If you're looking to partner with experts in this space, get in touch today and let us help you! #intelia #GenAI #artificialintelligence #machinelearning https://lnkd.in/giVMKRhk

👇 Follow me for more interesting contents. 👇 GitHub: 👉 https://lnkd.in/e7hZvG5Q LinkedIn:👉 https://lnkd.in/ej-jHj9s Google Scholar: 👉 https://lnkd.in/eQM_qKw7 #computervision #technologie #development #university #share #robotics #ai #google #datascience #data #datascientist #AI #ML #ArtificialIntelligence #MachineLearning #modeldeployment #reinforcementlearning #objectdetection #bourse #recherche #workshop #reinforcementlearning #SQL #nosql #sqlite #deeplearning

New article! Changbo Zhang, Hua Liao, and Jinhua Meng compare the performance of gaze interaction for map target selection, #gisChat, https://lnkd.in/ek2j-tj7 The eye tracking approach has the potential to facilitate more tailored and intelligent geoapplication interactions. There is a lack of knowledge regarding how gaze interaction influences user performance (accuracy and efficiency), visual behavior, and cognitive workload in real-world map-reading settings. In this study, we adopted a low-cost eye tracker and dwell time method to implement a fundamental gaze-based map interaction: target (point, line and area) selection. We administered a comparative experiment in which 36 participants were asked to perform a series of map target selection tasks using gaze-based and mouse interactions. The results demonstrated that gaze-based map interaction is comparable to mouse in terms of overall efficiency.

Demo from our eng Mert Gerdan: showing the 600ms end-to-end response times with the new Conversational Video Interface by Tavus. Aka the time it takes for the digital twin to respond to a human in a conversation 🤓