Hugging Face

Need embeddings for building multimodal LLMs, GenAI that spans more than just image/video, or a proper embedding model for curation purposes? Perhaps our 🤗 HuggingFace collection has something to offer. We have released Encords latest embedding model on top of our newly released dataset E-MM1. It supports embedding point clouds, images, videos, audio, and text.. and the best part, all the embeddings line up nicely so you can do cross-modal retrieval, embedding arithmetics, and all that stuff you'd expect. It's designed to be easy to use and very performant for it's size ⚡ Paper, models, datasets all available here: https://lnkd.in/dt-jQiiW Have fun!

Building datasets to train smaller, task-focused models used to be incredibly time-consuming. Very excited to see AI at Meta's SAM3 massively lower that barrier. Describe the class you want to detect and get annotated datasets automatically! Example: I just created a dataset of photographs from historical newspapers using the Europeana newspaper collection in a single command on Hugging Face Jobs. How it works: - Point the script to your image dataset - Choose the class to detect - Specify the output dataset for storing results You could then either directly use these results, do some extra annotation or filtering or use the dataset to fine-tune a small object detection model focused on your domain and task! hf jobs uv run \  --flavor a100-large \  -s HF_TOKEN=HF_TOKEN \  url-to-script/detect-objects.py \  -- davanstrien/newspapers-with-images-after-photography-big \  davanstrien/newspapers-photo-predictions \  --class-name "photograph" \  --confidence-threshold 0.4 Browse results: https://lnkd.in/ew5BEQkq Try it on your own dataset: https://lnkd.in/edWsyQDB

I’ve joined Hugging Face as a Developer Advocate! 🤗🚀 Super excited about this. I’ll be helping developers build AI applications with open models, sharing educational content, and supporting the community however I can. If you have ideas for videos/courses, feel free to reach out! Excited for the journey ahead! 🧡

The LLM space is blessed to have awesome tech reports that help push the field further exponentially. The diffusion community kinda lacks that a bit. Good folks at Photoroom decided to change that by releasing PRX under Apache 2.0 with solid reporting. Being Apache 2.0, PRX allows the community to improve it without limitations. Sure, it has its flaws, but that doesn't mean it cannot be improved. The pre-trained base models come with different flavors. They are not yet SOTA, but the point lies in the openness -- what does it take to train diffusion-based T2I models with modern recipes? The checkpoints released so far can be considered "base". The team is busy post-training to improve in different aspects. So, that's definitely something to look forward to! Check out the progress made so far: https://lnkd.in/gRfRJpZF

🚀 Slush Week is ON! Because community is what matters most, we’re launching a dedicated Networking Area to help founders, builders, investors and open-source enthusiasts connect alongside our side event « The power of open source : Building giants in the open » ! Alongside our new Networking Area, you’ll find a dedicated Expert Zone, where teams from Google DeepMind, Vercel and Le Robot will be sharing insights, demos, and answering your questions live. Who’s joining us at Slush this week? 🙌 https://lnkd.in/eWvYPDu8

Join us LIVE at MCP's first Birthday Bash kicking off at 10 am PT today!🎂 Don't miss out on details about the celebration from the co-hosts, Gradio and Anthropic AI. 🔥 We've also got an exciting lineup of speakers from Hugging Face, OpenAI, Google DeepMind, Modal, Blaxel (YC X25) AI , SambaNova AI, and Nebius Token Factory ready to share their insights. Catch us LIVE on YouTube here: https://lnkd.in/g8u2cGtk

cutting-edge computer vision pipelines often combine task-specific heads with backbones⚡️ but it can get messy loading them together for training, so Hugging Face transformers got you covered! 🤗 learn to combine DINOv3 with DETR for bleeding edge detection with Backbone API ⤵️ link to tutorial in comments 🙌🏻🫡