Scene Parsing Using Region-Based Generative Models

Abstract

Abstract-Semantic scene classification is a challenging problem in computer vision. In contrast to the common approach of using low-level features computed from the whole scene, we propose "scene parsing" utilizing semantic object detectors (e.g., sky, foliage, and pavement) and region-based scene-configuration models. Because semantic detectors are faulty in practice, it is critical to develop a region-based generative model of outdoor scenes based on characteristic objects in the scene and spatial relationships between them. Since a fully connected scene configuration model is intractable, we chose to model pairwise relationships between regions and estimate scene probabilities using loopy belief propagation on a factor graph. We demonstrate the promise of this approach on a set of over 2000 outdoor photographs, comparing it with existing discriminative approaches and those using low-level features.

Using Semantic Features for Scene Classification: how Good do they Need to Be?

“…We simulated a variety of imperfect detectors by randomly perturbing the manually labeled regions in D1 (details in [3]). We set the detection rates of individual detectors on each true material (both true and false positive rates) by counting performance of the corresponding actual detectors on a validation set.…”

An Interactive Approach for Filtering Out Junk Images From Keyword-Based Google Search Results

“…To achieve more accurate representation of the diverse visual properties of the images, both the global visual features and the local visual features are extracted for image content representation and similarity characterization [14][15][16][17][18][19][20]. To reduce the computational cost for feature extraction, we use the thumbnails from Google Images instead of the original-size images for feature extraction.…”

Image classification: Are rule-based systems effective when classes are fixed and known?

“…Clearly, if we can arrive at a set of rules that can achieve a high classification rate for these fixed classes, we can assert the feasibility of rule-based systems. Previous studies on scene classification have concentrated on modeling scenes or devising novel features like bag of words and its derivatives followed by machine learning [1,2,[6][7][8][9][10]. We show that a rule-based system can be simple and efficient.…”