EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Lobsters·Aug 18·5 min read
Using the railway network as a flatbed scanner
TL;DR
A novel digital scanning back for large format cameras was developed by moving the entire camera instead of the subject, allowing for high-resolution image capture. This approach simplifies construction and reduces costs compared to traditional methods.
✦ Why It Matters
Consider building a moving camera setup to capture high-resolution images without expensive scanning backs.
Key Takeaways
- Explore using moving cameras for unique imaging techniques.
- Consider low-cost alternatives to expensive scanning backs.
- Utilize accelerometer data for motion correction in images.
- Test different environments to optimize image capture.
How It Works
The technique involves moving a camera past a stationary subject to capture vertical lines of the scene, which are then stitched together to form a complete image. By using a high-speed camera like the Basler ruL2048-19gm, which captures nearly 19,000 lines per second, the system can gather more data than traditional methods.
Accelerometer data is integrated to correct for motion distortions, ensuring that the final image appears more natural.