We created an open-source computer vision model called Franca, which can perform as well or even better than some of the most advanced models currently available. This could lead to more reliable and accurate results in various applications, such as image recognition and object detection, and help pave the way for more transparent and reproducible AI research.
Read Full Paper →Scientists developed a new way to make diffusion models produce desired results, such as images or text with specific properties, without needing expensive training. This method, called Feynman-Kac steering, can improve sample quality and allow for more control over what's generated, making it easier to get the results you want from these models.
Read Paper →Researchers created a system to automatically generate high-quality images based on text instructions. This system can produce thousands of accurate image-editing examples without human help, allowing for faster and more efficient development of advanced image editing tools.
Read Paper →Scientists developed a new way to analyze medical images from CT scans by breaking them down into smaller parts, focusing on areas where doctors need more information. This approach allows for more accurate diagnoses and better treatment planning, with significant improvements over existing methods in detecting kidney problems, lung cancer, and COVID-19.
Read Paper →Scientists developed a new way to make computer simulations of human motion more accurate by combining artificial intelligence with text-based descriptions of tasks, allowing for better evaluation of worker behavior and safety in industrial settings. This improved method could help reduce errors and improve productivity in workplaces that rely on manual labor, making it a valuable tool for industries like manufacturing and construction.
Read Paper →
