activescott's Notes

CVAT Community is the free, self-hosted open-source edition of CVAT — one of the most widely used data annotation platforms for building high-quality visual datasets for computer vision and visual AI. Since 2018, CVAT has become one of the best-known data annotation tools in computer vision, with a large open-source community, millions of Docker pulls, and broad adoption across research and production AI teams.

Segment Anything Model 2 (SAM 2) is a foundation model towards solving promptable visual segmentation in images and videos. We extend SAM to video by considering images as a video with a single frame. The model design is a simple transformer architecture with streaming memory for real-time video processing. We build a model-in-the-loop data engine, which improves model and data via user interaction, to collect our SA-V dataset, the largest video segmentation dataset to date. SAM 2 trained on our data provides strong performance across a wide range of tasks and visual domains.

Buried inside the KIDS Act are provisions that will push online services to verify all users’ ages, require government-directed moderation policies for online speech, and even create new rules about private and encrypted communications. While supporters continue to claim this bill protects minors online, its requirements come at the expense of privacy, free expression, and the ability of people of all ages to use the internet without revealing sensitive data.

Turn raw visual data into model-ready datasets with CVAT annotation platform. Run it on your own infrastructure, CVAT cloud, or let our labeling team do the work.

You program the OpenMV Cam in Python. We make it easy to run machine vision algorithms and AI models on what the OpenMV Cam sees and then actuate hardware in the real world. Sense, plan, and act all in one Python script.

What makes microcontrollers unique is their low power consumption, low heat generation, small size, and ability to draw microwatts of power in deep sleep. This enables you to build tiny devices that can survive for years on batteries.

Beyond putting all these features into such a small footprint, we believe in giving you the tools to easily integrate the OpenMV Cam into any system. Each board exposes plenty of GPIO pins that provide SPI, I2C, I3C, UART, CAN, PWM, and ADC functionality.

For professionals, our schematics are available online so you can fully understand every OpenMV Cam and its accessories. You can modify and compile our firmware from GitHub, and SWD and JTAG are exposed for you to single step and debug your changes.

Vision language models are broadly defined as multimodal models that can learn from images and text. They are a type of generative models that take image and text inputs, and generate text outputs. Large vision language models have good zero-shot capabilities, generalize well, and can work with many types of images, including documents, web pages, and more. The use cases include chatting about images, image recognition via instructions, visual question answering, document understanding, image captioning, and others. Some vision language models can also capture spatial properties in an image. These models can output bounding boxes or segmentation masks when prompted to detect or segment a particular subject, or they can localize different entities or answer questions about their relative or absolute positions.

This is an informative, safe, comfortable way to begin welding. We supply everything needed.

For best results, Take this class, Mig 1 welding for noobs, then take either our Boot Camp (three weeknights) or Welding 101 which is a series of three classes with one built-in make up date. This class is offered as an intro so you can try welding out, but there is no reason not to take the boot camp or 101 series first, aside from it being a larger commitment.

Robostral Navigate is an 8B model that enables robots to autonomously navigate complex environments using only a single RGB camera, achieving 76.6% success on unseen R2R-CE benchmarks—outperforming multi-sensor approaches while being more efficient. Built entirely in-house with simulated data and token-efficient techniques, it generalizes across robot types and adapts to real-world obstacles unseen during training. The model combines pointing-based navigation with reinforcement learning for continuous improvement, paving the way for unified embodied AI in robotics.

State-of-the-art performance on R2R-CE

79.4% Success Rate on validation seen

76.6% Success Rate on validation unseen 

Operates from a single RGB camera, with no LiDAR or depth sensors

8B model, built in-house and trained entirely in simulation

Runs on wheeled, legged, and flying robots, and generalizes across robot sizes

Robust to differences in camera intrinsics

Token-efficient training via prefix-caching

A key ingredient of Robostral Navigate is an efficient training algorithm based on prefix-caching. Using a tree-based attention-masking strategy, our method compresses an entire episode into a single sequence, enabling training on all time steps in a single forward pass while preventing information leakage between time steps.

Compared to training with one sample per time step, our approach reduces the number of training tokens by 22× while preserving all of the learning signals. In practice, this method transforms training runs that would take months into runs that complete in days.

Parse Platform is your complete backend solution for mobile and web applications. Deploy anywhere, scale infinitely, own your data.

Most Fed watchers and financial analysts, though, don’t see evidence yet of broad labor productivity gains from AI that could justify a reduction in the Fed’s borrowing rates. Fed policymakers last proceeded with a quarter-point cut at their December meeting and hit pause ever since.

Instead, they’re zeroing in on the rapid data center buildout that’s swallowing up an enormous amount of storage and memory chips — critical components in smartphones, video game consoles, cars, and more — as a culprit for another wave of inflation. The gobbling frenzy weakens the case for lower interest rates further

Home chores robot.

“Go out and buy a Dell computer,” Trump said. It wasn’t the first time he’s made that pitch. He did so in May when lauding Dell CEO Michael Dell and his wife, Susan, who have pledged to donate more than $6 billion to the “Trump Accounts,” program, which launched July 4.

Trump, who has disclosed actively trading Dell shares, said he wants the couple to make back the money they’re giving to the accounts.

“We’re going to get him that money back one way or the other — and then I’ll ask for another $6 billion ... We’ll start the whole process all over again,” Trump said.

UN human rights group demands immediate release of Gaza doctor

A UN human ⁠rights ⁠body has called Israel’s detention of Gaza Dr ⁠Hussam Abu Safiya arbitrary and seeks his immediate release ⁠as rights groups and his lawyer warn his life is in imminent danger.

In its finding, ‌the UN Working Group on Arbitrary Detention said Israel’s actions contravened multiple articles of the Universal Declaration of Human Rights, as well ⁠as the International Covenant ⁠on Civil and Political Rights.

Over 90,000 houses damaged by Israeli attacks in southern Lebanon

Continued Israeli attacks have prevented more than 600,000 people from returning to southern Lebanon.

Israel is posting videos of its military blowing up neighbourhoods.

The soldiers have been doing this for months – wiping villages off the map – actions that rights groups say violate international law, reports Al Jazeera’s Zeina Khodr.