SalFoM: Dynamic Saliency Prediction with Video Foundation Models
This work addresses the challenge of adapting image foundation models to video for improved saliency prediction, which is incremental as it builds on existing video foundation model concepts.
The authors tackled the problem of video saliency prediction by introducing SalFoM, a novel encoder-decoder video transformer architecture based on video foundation models, which demonstrated superiority over state-of-the-art methods on benchmark datasets like DHF1K, Hollywood-2, and UCF-Sports.
Recent advancements in video saliency prediction (VSP) have shown promising performance compared to the human visual system, whose emulation is the primary goal of VSP. However, current state-of-the-art models employ spatio-temporal transformers trained on limited amounts of data, hindering generalizability adaptation to downstream tasks. The benefits of vision foundation models present a potential solution to improve the VSP process. However, adapting image foundation models to the video domain presents significant challenges in modeling scene dynamics and capturing temporal information. To address these challenges, and as the first initiative to design a VSP model based on video foundation models, we introduce SalFoM, a novel encoder-decoder video transformer architecture. Our model employs UnMasked Teacher (UMT) as feature extractor and presents a heterogeneous decoder which features a locality-aware spatio-temporal transformer and integrates local and global spatio-temporal information from various perspectives to produce the final saliency map. Our qualitative and quantitative experiments on the challenging VSP benchmark datasets of DHF1K, Hollywood-2 and UCF-Sports demonstrate the superiority of our proposed model in comparison with the state-of-the-art methods.