Edgar

new #kornia release out! check the new cool features! #computervision #opensource #artificialintelligence #innovation #pytorch #technology

@xtbot @Ken_Goldberg Pretty cool: I’ve been doing also something similar but instead of using openclaw, vibe coded a custom agent github.com/kornia/bubbalo…

I asked Codex to set up ROS middleware, configure a CSI camera, benchmark Gemma 4 models on my Jetson Orin Nano, adapt an OpenClaw-style runtime for VLM + reasoning (“Robotclaw” as I call it), and even build an iOS app to stream LiDAR, camera, GPS, and IMU data from an old iPhone to my robot rover. It’s honestly wild how capable these coding agents are now, and how much time they save. I even “write” way more tests now because the marginal cost is so low.

Here's Will's article on Agentic Coding for robotics: drive.google.com/file/d/1vNfgjq…

@spillai @chris_j_paxton @jaiselsingh Algorithms libraries are another story, probably that applies only for now to gluecode and middleware frameworks. Said that, I’m getting recently really amazing performance results with kernel optimisation with kornia-rs beating popular libraries

@edgarriba @chris_j_paxton @jaiselsingh Definitely. I keep thinking about this. At what abstraction do we think that a vibe-coded alternative doesn’t make sense? For example, would you consider vibe-coding cv2/libjpeg/pyav etc.

Again robotics software sucks. Its so bad. Part of what makes it so bad -- and what makes stuft like ROS useful -- is the amount of boring middleware code you need to write between sensors, processes, etc. Coding models make it so much easier to work with, its wonderful.

@bartdecrem You burned all the tokens

Project MacClaude is back on track

@chris_j_paxton @jaiselsingh The more vibe coding evolves, the more developers tools (including viz tools) or saas platforms will be easier to replicate/customise. Winners are low level libraries talking directly to metal and big platforms

@jaiselsingh I think these days with rust and python libraries, especially with visuals from rerun, you can vibe code your own "ros" easily enough

@karpathy Our team has been using similar ideas before each experiment to autotune all the parameters to also maximise gpu usage

I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the prompt (.md) - the AI agent iterates on the training code (.py) The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc. github.com/karpathy/autor… Part code, part sci-fi, and a pinch of psychosis :)

@openclaw is becoming the new Linux

@jiayuan_jy Make a karpathy swarm

Karpathy Guidelines for coding agents github.com/forrestchang/a…

@levelsio I have clawd running on a jetson orin nano

Why people run Clawdbot on a Mac Mini when they can run it on a VPS? Is it because you can do more regular ppl stuff on it cuz GUI?

The Overlooked Value of Test-time Reference Sets in Visual Place Recognition Mubariz Zaffar, Liangliang Nan, Sebastian Scherer, Julian F. P. Kooij tl;dr: if you have db (reference) images for VPR, finetune your model on it. arxiv.org/abs/2510.03751

#kornia used for the data augmentation

🚀 Welcome to the first #kornia virtual Performance Optimization Hackathon! 🗓 Dates: now – August 10, 2025 (AoE) 🏆 Prize Pool: Nintendo Switch 2 for Grand Prize Winner 🤝 Sponsored By: @codeflashAI 🧠 Hosted By: @kornia_foss You can still register docs.google.com/forms/d/e/1FAI…

@lukas_m_ziegler @farm_ng_inc @IlirAliu_

Build robots for a better future! 🚜 The @farm_ng_inc Amiga is more than just a farm robot — it’s an open-source development platform for agricultural innovation. Developers can easily build on top of its modular design, and developer-friendly SDK. Let's review other projects that were based on the Amiga robot! → At the Alcorn State University researchers developed a robot that uses UVC light to kill crops, pests, and diseases instead of chemical pesticides. It moves through crop rows using GPS and sensors to apply light precisely without harming plants. Such a cool one! → Another great project at Carnegie Mellon University, where they have developed a robot that plants trees automatically using a three-step process—digging, placing seedlings, and tamping the soil. It uses AI and sensors to navigate and ensure successful planting. → Last but not least, at Penn State University they created a weed-removing robot. This robot detects and removes weeds in high-tunnel farms. It uses cameras to map weed locations and applies precise herbicide amounts. This reduces chemical use, and labor, and improves crop management. As you can see, potential applications in agriculture are plentiful. I encourage you - let's build robots for better and more sustainable farms! Developers unite and build together!

Hey #opensource maintainers, we are trying to find some tools for #github to detect and score AI generated code in our repositories in @kornia_foss. Anyone knows or has experienced with any tool ?

@osanseviero kornia-rs 🦀

We're collaborating with open source maintainers to make sure Gemma models are well supported in day-0 when a new model is out! For Gemma 3: ✅HF transformers/bitsandbytes ✅Ollama ✅VLLM ✅llama.cpp ✅Unsloth ✅MLX ✅TGI ✅ONNX with transformers.js Who are we missing?

Happy to announce that @kornia_foss has been selected as mentoring org for the Google Summer of Code 2025 !!! @GoogleOSS I encourage everyone interested in the topics of #computervision #ai #robotics and #rust to apply

VNS is GO for Descent! Space-ng has developed the Vision Navigation Software (VNS) for Firefly Aerospace’s Blue Ghost lunar lander. After 4 years of design, development, test, integration and operations, our subsystem is about to support a mission-critical function: providing terrain relative navigation and hazard detection during Blue Ghost’s descent and landing on March 2, 2025. 🧵(1/4)

@XingyiHe1 @ducha_aiki Awesome. Looking forward to it. Would be great if we can get a ping once that happens. The community will appreciate it

@edgarriba @ducha_aiki Hi Edgar, thanks for your interest in our work! We are glad that our model can be considered included in kornia. The models and code will be open-sourced later. We can convert them (our pretrained ELoFTR and ROMA models) to onnx if needed.

Excited to share our work MatchAnything: We pre-train strong universal image matching models that exhibit remarkable generalizability on unseen multi-modality matching and registration tasks. Project page: zju3dv.github.io/MatchAnything/ Huggingface Demo: huggingface.co/spaces/LittleF…

new releases are out. And more to come! Join our discord: discord.gg/p5MrYyNd

📢 New Year Releases are available !! 🦀 kornia-rs v0.1.8: github.com/kornia/kornia-… 🐍 kornia v0.8.0: github.com/kornia/kornia/… 👉Join our Discord: discord.gg/p5MrYyNd #computervision #pytorch #rust #opensource #geometry #robotics #deeplearning