SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2303.09941 · ICCV · 2023

Leaping Into Memories: Space-Time Deep Feature Synthesis

Alexandros Stergiou

arXiv · PDF · Open in the Atlas

Code that ran

We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Illustration of our proposed LEAPS model inversion. Given a video input x * initialized with noise, we use a stimulus video v of target class y, to prime video model Φ(•). We iteratively optimize x * to synthesize and visually represent the internal spatiotemporal features of Φ(•). To impose diversity over the synthesized videos, feature statistics from a domain-specific verifier network S(•) are distilled to regularize x * . To preserve the continuity of motions across frames, x * is temporally regularized at each training iteration. Although the illustration visualizes internal representations as C ×T ×H ×W volumes, they can also be flattened patches CP 3 × T HW P 3 of P 3 resolution as in Vision Transformers.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2303.09941")
get_code_for_paper("2303.09941")
have("2303.09941")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2303.09941.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2303.09941.svg)](https://syntology.ai/paper/2303.09941)

Connect an agent — have() is free.