Loading video...

Video Failed to Load

Go Home

A new open source computer vision model arrives. Plainsight's founder, Kit Merker joins me to explain why it's important for developers building for companies. Grok, below, tells you what you will learn by watching.

13,160 views • 1 year ago •via X (Twitter)

5 Comments

Robert Scoble's profile picture
Robert Scoble1 year ago

Here is what @Grok says you will learn: Unlock the Future of Computer Vision with Open Filter: Key Takeaways from an Exclusive Interview! 🌐 Are you a developer curious about how to harness the power of computer vision for your next big project? Or maybe you’re just fascinated by the explosion of camera-driven tech in AR glasses, robots, drones, and beyond? I recently dove into an in-depth interview with Kit, CEO of Plain Sight, where they unveiled Open Filter, a game-changing open-source project set to democratize computer vision. Here’s a detailed breakdown of what you’ll learn by watching this insightful conversation—trust me, it’s packed with ideas that could spark your next innovation! 📷1. What is Open Filter, and Why Does It Matter? You’ll discover Open Filter, a new open-source framework (Apache 2.0 licensed) designed to make computer vision accessible to developers of all skill levels. Kim introduces the concept of “filters”—think Instagram filters, but for AI-powered vision apps. These filters combine machine learning models, custom code, and a universal runtime to create composable, scalable vision pipelines. Whether you’re building apps for AR glasses, warehouse drones, or traffic cameras, Open Filter simplifies the process. You’ll learn: How filters work as a universal abstraction for vision tasks. Why Plain Sight is open-sourcing this tech to set a “de facto standard” for vision apps. The big picture: how Open Filter fits into the emerging Vision Internet, where cameras and AI converge to process data at scale. 2. Real-World Applications for Developers The interview is a goldmine of practical examples showing how developers can use Open Filter. Kim walks through a license plate reading demo, breaking down how three filters (detection, cropping, and OCR) work together to process video frames. You’ll learn how to: Build vision apps by chaining filters for tasks like object detection, text recognition, or image cropping. Optimize performance with utilities like frame deduplication to save GPU resources. Apply Open Filter to diverse use cases, from: Consumer tech: AR glasses that translate menus or find lost keys. Industrial settings: Inventory tracking in warehouses or quality control in factories. Public safety: Detecting anomalies like drunk drivers via traffic cameras. Robotics: Enabling robots to recognize objects like a Coke can or a Tide bottle. 3. Scaling Vision Apps: From Prototype to Production If you’ve ever built a computer vision system, you know scaling is a nightmare. Kit explains why bespoke systems (like those in drone-based inventory control) hit walls when you add more cameras, models, or features. You’ll learn how Open Filter solves this by: Treating vision apps like microservices, making them modular and reusable. Allowing you to swap models (e.g., from Hugging Face or custom-trained) without rewriting pipelines. Managing the data lifecycle, from collecting raw camera data to annotating and training models. Supporting deployment across cloud, edge, and devices (e.g., glasses, robots) with its Python-based, write-once-run-anywhere design. For developers with existing systems, you’ll see why you don’t need to “rip and replace”—Open Filter lets you gradually adopt its composable approach. 4. Navigating Trade-Offs in Vision Workloads The interview dives into the diverse needs of vision apps, from power-constrained AR glasses to latency-critical traffic systems. You’ll learn how developers can use Open Filter to balance trade-offs like: Power efficiency: Optimizing for battery life on glasses or robots with smaller models or fewer frames. Latency: Processing high-frame-rate video in real-time for safety applications. Cost: Reducing GPU usage with smart filtering to make vision affordable at scale. Kim’s insights will help you think strategically about architecting vision apps for specific hardware and use cases. 5. The Vision Internet and AI Agents Kim’s concept of the Vision Internet is a highlight—a future where cameras generate data for AI, not humans, akin to “Netflix for robots.” You’ll learn: Why vision infrastructure needs to prioritize AI consumption (e.g., JSON blobs, alerts) over smooth video playback. How Open Filter enables AI agents to process camera feeds dynamically, e.g., monitoring warehouse inventory and triggering alerts only for anomalies. The parallels with cloud computing’s evolution (monoliths to microservices) and how Open Filter could be the “Kubernetes” of vision workloads. This vision will inspire you to rethink how cameras and AI can transform industries like retail, logistics, and manufacturing. 6. Data Pipelines and Model Training Beyond inference, Open Filter shines in managing the data supply chain. You’ll learn how it supports: Annotation pipelines: Deduplicating frames, pre-annotating images, and sending only tough cases to human annotators (a $5B industry!). Model training: Using real-world data (e.g., nightclub footage) to train specialized models, like detecting Gucci shoes. Continuous improvement: Creating feedback loops where data from cameras refines models over time. This is crucial for developers building apps that need to evolve with new data or requirements. 7. Business Model and Community You’ll get the lowdown on how Plain Sight is approaching Open Filter: Free forever: The open-source project is free for academic, research, or prototyping use. Paid support: Plain Sight offers commercial services for production-grade deployments (details at Community-driven: A Discord server, office hours, and example filters make it easy to get started and collaborate. Kim’s goal is to make filters a standard, and you’ll see how the community can shape its future. 8. Privacy and Ethical Considerations The interview touches on hot-button issues like face recognition and surveillance. You’ll learn: Open Filter is use-case agnostic, enabling apps from badge reading at conferences to inventory tracking, but developers must handle privacy compliance. Why models for tasks like face recognition require separate datasets and systems, not just Open Filter. How social norms and software controls (e.g., on-device processing) will shape the adoption of vision tech in consumer devices like glasses or robots. 9. Demo and Practical Insights Kim shares a live demo of Open Filter reading a license plate, showing how simple it is to set up a pipeline. You’ll see: The code and filters in action, processing video frames in real-time. How developers can use the Open Filter API with tools like Claude or Cursor for AI-assisted coding. Tips for prototyping and scaling apps using off-the-shelf models or custom ones. 10. What’s Next for Open Filter? The interview hints at Open Filter’s future, including: Multimodal support: Expanding to audio, lidar, and geospatial data. Agent integration: Powering vision-enabled AI agents for enterprise systems like Salesforce or SAP. Community growth: Encouraging developers to contribute filters and ideas via GitHub and Discord. You’ll leave inspired to join the movement and explore how Open Filter can accelerate your projects. Why Watch This Interview? This isn’t just a tech talk—it’s a window into the future of computer vision, where cameras and AI are reshaping how we interact with the world. Whether you’re a seasoned developer or just starting out, you’ll gain: Practical skills: Learn how to build and scale vision apps with Open Filter. Big-picture vision: Understand the Vision Internet and its impact on industries. Inspiration: See how you can innovate in consumer tech, robotics, or industrial automation. Where to Watch and Learn More Check out the full interview to see Kit's demo and hear their passion for democratizing vision tech! Then, dive into Open Filter: 📷 Website: for GitHub, docs, and community links. 📷 : for production-grade help. 📷 Co🤝n3" style="color: black; background-color: transparent; font-family: sans-serif;">: Join the Discord or attend office hours to connect with other developers. Call to Action Watch the interview, explore Open Filter, and share your thoughts! What vision apps are you dreaming up? Could Open Filter help you build an app for AR glasses, warehouse drones, or even a nightclub crowd analyzer? Drop your ideas below, and let’s spark a conversation about the future of computer vision! 📷 🚀

Robert Scoble's profile picture
Robert Scoble1 year ago

And here is a list of 1,400 in Computer Vision if you want to watch the computer vision community:

UserInterface's profile picture
UserInterface4 years ago

Let's talk about building businesses, #investing, #entrepreneurship & #crypto.

🇺🇸Jacques Boddy🇺🇸's profile picture
🇺🇸Jacques Boddy🇺🇸1 year ago

@PlainsightAI @KitMerker Awesomeness

Joe Ai Wild's profile picture
Joe Ai Wild1 year ago

@PlainsightAI @KitMerker Love the Grok analysis, such a good wittle groky poo

Related Videos