Extreme Amodal Face Detection

2025-10-08

AI Summary

This work formally defines and investigates the novel task of single-image extreme non-modal face detection: localizing faces in an input image that are partially or even fully occluded yet lie within the extended field of view, using only contextual cues from a single frame. Unlike existing approaches relying on multi-frame interpolation or generative sampling, we propose an efficient, sampling-free autoregressive heatmap detection framework. It employs a selective coarse-to-fine decoding strategy that leverages local contextual features to model the global face distribution. Crucially, our method requires neither generative models nor video sequences. Evaluated on a newly constructed benchmark, it significantly outperforms both multi-frame and generative baselines, demonstrating the feasibility and effectiveness of extreme non-modal reasoning from a single image.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Extreme amodal detection is the task of inferring the 2D location of objects that are not fully visible in the input image but are visible within an expanded field-of-view. This differs from amodal detection, where the object is partially visible within the input image, but is occluded. In this paper, we consider the sub-problem of face detection, since this class provides motivating applications involving safety and privacy, but do not tailor our method specifically to this class. Existing approaches rely on image sequences so that missing detections may be interpolated from surrounding frames or make use of generative models to sample possible completions. In contrast, we consider the single-image task and propose a more efficient, sample-free approach that makes use of the contextual cues from the image to infer the presence of unseen faces. We design a heatmap-based extreme amodal object detector that addresses the problem of efficiently predicting a lot (the out-of-frame region) from a little (the image) with a selective coarse-to-fine decoder. Our method establishes strong results for this new task, even outperforming less efficient generative approaches.

Problem

• Detecting faces outside image frames using contextual cues
• Predicting unseen face locations from single image input
• Developing efficient sample-free amodal face detection method

Innovation

• Single-image extreme amodal face detection
• Contextual cues infer unseen faces
• Coarse-to-fine heatmap-based detector