FGOS-Net framework animation

This directory contains an eight-second loop derived directly from the paper’s original framework figure. Module names, formulas, arrows, relative layout, and colors remain raster-faithful because every animated region is cropped from the source image rather than redrawn or retyped.

Outputs

• fgos-net-framework-loop.mp4 — 1280×720, 30 fps, H.264, 8 seconds
• fgos-net-framework-loop.gif — 960×540, 12 fps, infinite loop
• fgos-net-framework-loop.svg — standalone animated SVG with the source image embedded
• eccv-2026-logo.png — transparent ECCV 2026 Malmö conference mark supplied for the animation
• render_fgos_animation.py — editable timing, crops, flow paths, and highlight definitions

Animation sequence

The 16:9 composition follows the restrained academic-teaser language of Chen Liu’s Dispersion demo: a persistent paper-title line, the ECCV mark in the upper right, a compact research-question banner, a white canvas, and progressive diagram construction with minimal color.

1. Input and stem enter from the left.
2. The FGOS encoder assembles hierarchically.
3. The GFA decoder and top-down fusion path appear.
4. The BRM head and FGOS Block detail enter.
5. The FA Block rises into place and the main connections are progressively drawn.
6. The untouched full framework resolves, then FA-Scan, ASGP, and BRM receive restrained highlights.
7. The complete framework holds before fading back to the faint initial state.

Regenerate

From the repository root:

python3 assets/animations/fgos-net/render_fgos_animation.py

Requirements: Python 3, Pillow, and FFmpeg with libx264 support.