Dynamic-DINO: Fine-Grained Mixture of Experts Tuning for Real-time Open-Vocabulary Object Detection
It addresses the challenge of efficient and accurate object detection in real-time applications, though it appears incremental as it builds on existing models like Grounding DINO.
This work tackles the problem of applying Mixture of Experts (MoE) architectures to real-time open-vocabulary object detection, proposing Dynamic-DINO, which extends Grounding DINO 1.5 Edge into a dynamic inference framework via MoE-Tuning and achieves better performance despite using less pretraining data.
The Mixture of Experts (MoE) architecture has excelled in Large Vision-Language Models (LVLMs), yet its potential in real-time open-vocabulary object detectors, which also leverage large-scale vision-language datasets but smaller models, remains unexplored. This work investigates this domain, revealing intriguing insights. In the shallow layers, experts tend to cooperate with diverse peers to expand the search space. While in the deeper layers, fixed collaborative structures emerge, where each expert maintains 2-3 fixed partners and distinct expert combinations are specialized in processing specific patterns. Concretely, we propose Dynamic-DINO, which extends Grounding DINO 1.5 Edge from a dense model to a dynamic inference framework via an efficient MoE-Tuning strategy. Additionally, we design a granularity decomposition mechanism to decompose the Feed-Forward Network (FFN) of base model into multiple smaller expert networks, expanding the subnet search space. To prevent performance degradation at the start of fine-tuning, we further propose a pre-trained weight allocation strategy for the experts, coupled with a specific router initialization. During inference, only the input-relevant experts are activated to form a compact subnet. Experiments show that, pretrained with merely 1.56M open-source data, Dynamic-DINO outperforms Grounding DINO 1.5 Edge, pretrained on the private Grounding20M dataset.