img2img-turbo
One-step image-to-image with Stable Diffusion turbo: sketch2image, day2night, and more
g
gaparmar
Professional software vendor delivering innovative solutions on the Softono platform. Specialized in both open-source and proprietary software development.
Total Products
2
Software by gaparmar
clean-fid
Clean FID is a PyTorch-based library for evaluating generative models by calculating the Frechet Inception Distance (FID) and Kernel Inception Distance (KID). It addresses critical inconsistencies found in standard FID implementations caused by improper low-level image resizing, quantization, and compression artifacts. The library implements mathematically correct resizing filters, such as PIL-bicubic and Lanczos, to ensure results are comparable across different research groups and publications. It also supports computing FID using CLIP features, offering an alternative evaluation metric. By standardizing preprocessing steps, Clean FID prevents large score discrepancies that arise from library-specific bugs or default settings in frameworks like Tensorflow, PyTorch, and OpenCV. This tool is essential for researchers needing reliable benchmarks for Generative Adversarial Networks and other image generation architectures. It was introduced in the CVPR 2022 paper On Aliased Resizing and Surprising Subtleties in