GRAINet: mapping grain size distributions in river beds from UAV images with convolutional neural networks

نویسندگان

چکیده

Abstract. Grain size analysis is the key to understand sediment dynamics of river systems. We propose GRAINet, a data-driven approach analyze grain distributions entire gravel bars based on georeferenced UAV images. A convolutional neural network trained regress as well characteristic mean diameter from raw GRAINet allows for holistic bars, resulting in (i) high-resolution estimates and maps spatial distribution at large scale (ii) robust grading curves bars. To collect an extensive training dataset 1491 samples, we introduce digital line sampling new annotation strategy. Our evaluation 25 along six different rivers Switzerland yields high accuracy: diameters have absolute error (MAE) 1.1 cm, with no bias. Robust can be extracted if representative data are available. At bar level MAE predicted even reduced 0.3 ranging 1.3 29.3 cm. Extensive experiments were carried out study quality generalization capability locations, model performance respect human labeling noise, limitations current model, potential images low resolutions.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Predicting Semantic Descriptions from Medical Images with Convolutional Neural Networks

Learning representative computational models from medical imaging data requires large training data sets. Often, voxel-level annotation is unfeasible for sufficient amounts of data. An alternative to manual annotation, is to use the enormous amount of knowledge encoded in imaging data and corresponding reports generated during clinical routine. Weakly supervised learning approaches can link vol...

متن کامل

Introducing a method for extracting features from facial images based on applying transformations to features obtained from convolutional neural networks

In pattern recognition, features are denoting some measurable characteristics of an observed phenomenon and feature extraction is the procedure of measuring these characteristics. A set of features can be expressed by a feature vector which is used as the input data of a system. An efficient feature extraction method can improve the performance of a machine learning system such as face recognit...

متن کامل

Using Convolutional Neural Networks to Analyze Function Properties from Images

We propose a system for determining properties of mathematical functions given an image of their graph representation. We demonstrate our approach for twodimensional graphs (curves of single variable functions) and three-dimensional graphs (surfaces of two variable functions), studying the properties of convexity and symmetry. Our method uses a Convolutional Neural Network which classifies func...

متن کامل

Tree Species Identification from Bark Images Using Convolutional Neural Networks

Tree species identification using images of the bark is a challenging problem that could help in tasks such as drone navigation in forest environment and autonomous forest inventory management. It also brings more value to harvesting operations as it leads to greater market values of trees. While the recent progress in deep learning showed its effectiveness for visual classification, it cannot ...

متن کامل

Deep Convolutional Neural Networks and Noisy Images

The presence of noise represent a relevant issue in image feature extraction and classification. In deep learning, representation is learned directly from the data and, therefore, the classification model is influenced by the quality of the input. However, the ability of deep convolutional neural networks to deal with images that have a different quality when compare to those used to train the ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: Hydrology and Earth System Sciences

سال: 2021

ISSN: ['1607-7938', '1027-5606']

DOI: https://doi.org/10.5194/hess-25-2567-2021