Raman Labs

Raman Labs also provides a variety of machine learning modules for developers, hobbyists and tinkerers to create their own creative projects. The modules let people easily add machine learning abilities to their projects with a few lines of code and ensure the results are fast and in real time.

Among the features:

• Speed: Runs in real time even with high-resolution data on consumer-grade CPUs.
• Simplicity: Has a clean and lightweight API to make integration easy.
• Robustness: Works well with large-scale and resolution variations, for example selfie videos and high-resolution video.
• Versatility: Runs on Python 3 and Numpy, so it can run on VPS, local machines or other infrastructure.
• Adaptability: Automatically adjusts to the available computing power of the system for optimal performance.

Raman Labs supports several platforms, including x86_64 architectures, and is available for integration with Python, but the company plans to support other languages through a C ABI-based SDK. The modules can be used for a variety of computer-vision tasks, including face detection, pose detection and face embedding.

The project also includes tools like Hachi, a self-hosted web app for natural language search in personal videos and images, and VideoIntel, a native app for real-time semantic search in videos. VideoIntel can perform tasks like recognizing multiple objects, generating hotspots and registering in-video.

For hardware, Raman Labs supports Intel/AMD 64-bit architectures with AVX2 instructions and NVIDIA/CUDA graphics cards. For software, it requires ONEDNN and OpenBLAS for CPU operations and CUDA TOOLKIT and CUDNN for GPU operations.

Raman Labs prioritizes user privacy, storing only necessary information like email and hashed passwords, with payment information handled by Level 1 PCI DSS Compliant Razorpay. The End-user License grants personal use on multiple devices, as long as it remains fully local.

Although Raman Labs is geared for personal use, its performance and flexibility means it can be used for other creative projects. You can check tutorials and product details on the company's website to learn more about the machine learning modules.

Published on June 14, 2024