Revisiting Single Image Depth Estimation: Toward Higher Resolution Maps with Accurate Object Boundaries
نویسندگان
چکیده
We revisit the problem of estimating depth of a scene from its single RGB image. Despite the recent success of deep learning based methods, we show that there is still room for improvement in two aspects by training a deep network consisting of two sub-networks; a base network for providing an initial depth estimate, and a refinement network for refining it. First, spatial resolution of the estimated depth maps can be improved using skip connections among the sub-networks which are trained in a sequential fashion. Second, we can improve estimation accuracy of boundaries of objects in scenes by employing the proposed loss functions using depth gradients. Experimental results show that the proposed network and methods improve depth estimation performance of baseline networks, particularly for reconstruction of small objects and refinement of distortion of edges, and outperform the state-of-the-art methods on benchmark datasets.
منابع مشابه
Object-Based Classification of UltraCamD Imagery for Identification of Tree Species in the Mixed Planted Forest
This study is a contribution to assess the high resolution digital aerial imagery for semi-automatic analysis of tree species identification. To maximize the benefit of such data, the object-based classification was conducted in a mixed forest plantation. Two subsets of an UltraCam D image were geometrically corrected using aero-triangulation method. Some appropriate transformations were perfor...
متن کاملDepth-Encoded Hough Voting for Joint Object Detection and Shape Recovery
Detecting objects, estimating their pose and recovering 3D shape information is a critical problem in many vision and robotics applications. This paper addresses the above needs by proposing a new method called DEHV Depth-Encoded Hough Voting detection scheme. Inspired by the Hough voting scheme introduced in [13], DEHV incorporates depth information into the process of learning distributions o...
متن کاملObject-Oriented Method for Automatic Extraction of Road from High Resolution Satellite Images
As the information carried in a high spatial resolution image is not represented by single pixels but by meaningful image objects, which include the association of multiple pixels and their mutual relations, the object based method has become one of the most commonly used strategies for the processing of high resolution imagery. This processing comprises two fundamental and critical steps towar...
متن کاملA Deep Model for Super-resolution Enhancement from a Single Image
This study presents a method to reconstruct a high-resolution image using a deep convolution neural network. We propose a deep model, entitled Deep Block Super Resolution (DBSR), by fusing the output features of a deep convolutional network and a shallow convolutional network. In this way, our model benefits from high frequency and low frequency features extracted from deep and shallow networks...
متن کاملMapping Impervious Surface Area Using High Resolution Imagery: a Comparison of Object-based and per Pixel Classification
Impervious surface area is a key indicator of environmental quality. Satellite remote sensing of impervious surface has focused on subpixel analysis via various forms of statistical estimation, subpixel classification, and spectral mixture analysis, using medium resolution Landsat TM or ETM+ data. Maps of impervious surface area from these studies provide useful inputs to planning and managemen...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2018