Academic
Publications
Region-based Segmentation and Object Detection

Region-based Segmentation and Object Detection,Stephen Gould,Tianshi Gao,Daphne Koller

Region-based Segmentation and Object Detection   (Citations: 10)
BibTex | RIS | RefWorks Download
Object detection and multi-class image segmentation are two closely related tasks that can be greatly improved when solved jointly by feeding information from one task to the other (10, 11). However, current state-of-th e-art models use a separate representation for each task making joint inferen ce clumsy and leaving the classification of many parts of the scene ambiguous. In this work, we propose a hierarchical region-based approa ch to joint object detection and image segmentation. Our approach simultaneously reasons about pixels, regions and objects in a coherent probabilistic mod el. Pixel appearance features allow us to perform well on classifying amorphous background classes, while the explicit representation of regions facilitate th e computation of more so- phisticated features necessary for object detection. Impo rtantly, our model gives a single unified description of the scene—we explain every pixel in the image and enforce global consistency between all random variables in our model. We run experiments on the challenging Street Scene dataset (2) and show signifi- cant improvement over state-of-the-art results for object detection accuracy.
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...Similar in spirit, but focusing on a complex graphical model that makes more use of context, is the work in [12]...

    Thomas Broxet al. Object segmentation by alignment of poselet activations to image conto...

    • ...To harness shape features, approaches such as [5, 14] have instead started with an initial segmentation and then refined these segments iteratively...
    • ...On the other hand, approaches such as [5, 14, 18] start from an imperfect segmentation and then refine it iteratively by optimizing a cost function defined on segments and appearance matchings...

    Xi Chenet al. Piecing together the segmentation jigsaw using context

    • ...Intuitively, jointly solving these two tasks can help each benefit from each other: a bounding box object detector can offer the segmentation model good location and shape priors [12, 22]; a segmentation model can prov ide the detector cleaner features, especially in the face of clutter [8]...
    • ...Recently, both Gould et al. [8] and Ladicky e t al. [13] c ombined an object detector with an image segmenter...

    Tianshi Gaoet al. A segmentation-aware object detection model with occlusion handling

    • ...Several solutions to this problem have been proposed using RF formulations [22, 16, 21, 20, 10, 3, 15, 5, 4, 18, 8, 2, 9]. These algorithms define a RF whose sites represent pixels in the image or superpixels obtained by oversegmentation of the image...
    • ...Related work. Recent work on JCaS has addressed the issue of using top-down object models that generalize to the case of multiple categories [10, 21, 5, 4, 23, 9]. Our framework differs from these works in that the object category information is encoded using the global BoF model...

    Dheeraj Singarajuet al. Using global bag of features models in random fields for joint categor...

    • ...There are many methods for using segmented superpixels and merging them together to form an object boundary [9, 12]...
    • ...Russell et al. [30] introduced the “soup of segments” idea, where multiple segmentations of an image are obtained, and then all the segments are considered together as building blocks in tasks such as object discovery [30], spatial support [25], or joint object classification and segmentation [12, 27]...

    Olga Russakovskyet al. A Steiner tree approach to efficient object detection

Sort by: