We’re thrilled to announce the release of Savant version 0.2.3, a cutting-edge computer vision and video analytics framework optimized for Nvidia hardware! Here’s what’s new:
1️⃣ Introducing new artist functionality, making it easier than ever to draw on frames without code. Our advanced syntax lets you declare borders, backgrounds, labels, dots, and blurring directly in the YAML configuration file. No more Python coding needed for customization!
2️⃣ Enjoy enhanced preprocessing capabilities, empowering you to modify object metadata and images effortlessly. Utilize OpenCV CUDA functionality for efficient cropping, rotation, and composition of images in GPU memory, eliminating the need for CPU RAM. A demo showcasing Age/Gender prediction is in the works for version 0.2.4!
3️⃣ A100 and GeForce users, rejoice! We’ve addressed GPU-specific considerations, such as limited NVENC resources on A100 and stream encoding limitations on GeForce. Savant now provides clear error messages when encountering these scenarios.
4️⃣ We’ve made significant improvements to the RTSP source adapter, ensuring more predictable performance. In our next release, we plan to switch the implementation from Gstreamer to FFmpeg, known for its superior RTSP support. Rust-based implementation of the FFmpeg source library is already in place.
5️⃣ Our demos have been updated to reflect the new artist functionality and other changes.
6️⃣ As part of our ongoing optimizations, we’re progressively transitioning Savant’s internals to Rust. This move eliminates the Global Interpreter Lock (GIL) and operations serialization in Python, leading to enhanced performance. The People Detection and Face blurring demo already benefits from these optimized operations.
7️⃣ We’ve also improved our documentation with additional information, ensuring a smoother experience for users.
🎯 Looking ahead to version 0.2.4, we have exciting plans:
1️⃣ An upcoming demo will showcase metadata and image preprocessing for Facial Age/Gender prediction.
2️⃣ Expect a universal FFmpeg-based source adapter that supports various sources like RTSP, RTMP, USB-cam, and more.
3️⃣ Several internal objects will be replaced with efficient Rust-based implementations.
4️⃣ Conditional artist and encoding functionality will be introduced, allowing streamlined selection of streams for object display and video encoding, optimizing hardware usage and processing efficiency.
🔗 Join our Discord and visit GitHub to learn more about Savant’s cutting-edge features and to stay updated with our latest advancements. Let’s shape the future of computer vision together!