CLIP-Guided Adaptable Self-Supervised Learning for Human-Centric Visual Tasks

TL;DR

CLASP is a novel unsupervised pre-training framework that leverages CLIP's multi-level semantic pseudo-labels and a prompt-controlled MoE module to improve human-centric visual task performance.

Contribution

This work introduces CLASP, combining CLIP-guided pseudo-labels with a dynamic MoE module for adaptable, multi-level feature learning in human-centric visual analysis.

Findings

Outperforms existing unsupervised pre-training methods on multiple benchmarks.

Effectively integrates multi-level semantic cues for richer representations.

Enhances transferability across diverse human-centric tasks.

Abstract

Human-centric visual analysis plays a pivotal role in diverse applications, including surveillance, healthcare, and human-computer interaction. With the emergence of large-scale unlabeled human image datasets, there is an increasing need for a general unsupervised pre-training model capable of supporting diverse human-centric downstream tasks. To achieve this goal, we propose CLASP (CLIP-guided Adaptable Self-suPervised learning), a novel framework designed for unsupervised pre-training in human-centric visual tasks. CLASP leverages the powerful vision-language model CLIP to generate both low-level (e.g., body parts) and high-level (e.g., attributes) semantic pseudo-labels. These multi-level semantic cues are then integrated into the learned visual representations, enriching their expressiveness and generalizability. Recognizing that different downstream tasks demand varying levels of semantic granularity, CLASP incorporates a Prompt-Controlled Mixture-of-Experts (MoE) module. MoE dynamically adapts feature extraction based on task-specific prompts, mitigating potential feature conflicts and enhancing transferability. Furthermore, CLASP employs a multi-task pre-training strategy, where part- and attribute-level pseudo-labels derived from CLIP guide the representation learning process. Extensive experiments across multiple benchmarks demonstrate that CLASP consistently outperforms existing unsupervised pre-training methods, advancing the field of human-centric visual analysis.