Loading…
A simple, client-side tool to visualize high-dimensional embeddings (e.g., image features) in an interactive 3D map using TSNE and UMAP. It is a fork of Google's TensorFlow Standalone Embedding Projector, updated to support UMAP and decoupled from specific ML libraries with a lightweight dependency stack. It uses a pretrained InceptionV3 feature extractor and outputs a static, publishable visualization that requires no server.
AI named TSNE-UMAP-Embedding-Visualisation from December 2025 to February 2026.