Cycle Consistency in Video Object-Centric Learning

2026-05-28

AI Summary

This work addresses the challenge of feature collapse in self-supervised video object-centric learning, where traditional explicit cycle consistency fails in implicit slot spaces due to the inherent randomness and ambiguity of scene decomposition. To overcome this limitation, the paper introduces Implicit Cycle Consistency (ICC), which reformulates cycle consistency as a constraint on the continuous manifold of video reconstruction rather than enforcing rigid alignment between slots. By promoting soft consensus among slots instead of strict correspondence, ICC effectively mitigates feature collapse. The proposed method achieves state-of-the-art performance on complex video object-centric learning benchmarks, significantly improving both object discovery and cross-frame association capabilities.

Abstract

Self-supervised video Object-Centric Learning (OCL) aims to discover distinct objects and associate them across time, whereas self-supervised Multi-Object Tracking (MOT) focuses on associating pre-defined object detections or segmentations. Although well-established in MOT, Cycle Consistency (CC) cannot naively or explicitly apply to the latent slot space of OCL. Unlike the deterministic and ideal object representations in MOT, OCL slots are inherently stochastic and ambiguous due to non-unique scene decompositions. Enforcing explicit cycle consistency (ECC) on slots imposes rigid mean seeking. This severely penalizes the model for exploring alternative but equally valid decompositions, thereby driving towards feature collapse. To resolve this dilemma, we propose \textit{Implicit Cycle Consistency (ICC)}, which shifts the cycle-consistency constraint from the restrictive slot space to the continuous reconstruction manifold, encouraging slots to reach a soft consensus on collectively interpreting the visual scene rather than forcing rigid point-to-point feature alignment. Extensive experiments on complex video OCL benchmarks demonstrate that ICC avoids feature collapse and outperforms ECC baselines. Our source code, model checkpoints and training logs are provided on https://github.com/Genera1Z/ICC.

Problem

• Cycle Consistency
• Object-Centric Learning
• Self-supervised Learning
• Feature Collapse
• Video Understanding

Innovation

• Implicit Cycle Consistency
• Object-Centric Learning
• Self-supervised Learning
• Video Understanding
• Slot Representation