Search papers, labs, and topics across Lattice.
This paper addresses the problem of feature collapse when applying cycle consistency to self-supervised video object-centric learning (OCL). They argue that explicit cycle consistency (ECC) in the latent slot space of OCL is too restrictive due to the stochastic nature of scene decomposition. To overcome this, they propose Implicit Cycle Consistency (ICC), which enforces cycle consistency on the reconstruction manifold instead of the slot space, leading to improved performance on video OCL benchmarks.
Stop forcing object-centric learning models into rigid feature alignment; cycle consistency works better when applied to the reconstructed visual scene.
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.
