Forest Species Recognition Using Deep Convolutional Neural Networks

Hafemann, Luiz G.; Oliveira, Luiz S.; Cavalin, Paulo Rodrigo

doi:10.1109/icpr.2014.199

Cited by 102 publications

(66 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We cannot compare to AlexNet as it requires input images of size 227x227. Our method outperforms the state of the art [8] on the Macroscopic forest species dataset (+1.4%) and obtains very similar results on the Microscopic one (-0.3%) while using a less complex approach. Table 6: Classification results (accuracy %) on kth-tips-2b using AlexNet and T-CNN-3 separately and combined as well as the state of the art method with a medium depth CNN (VGG-M).…”

Section: Results On Larger Imagesmentioning

confidence: 68%

“…More recently, Hafemann [8] applied CNN to a forest species classification, similar to a texture classification problem. While more complex and more accurate than [22], this approach still does not take the characteristics of texture images (statistical properties and repeated patterns) into consideration as it is a simple application of a standard CNN to a texture dataset.…”

Section: Related Workmentioning

confidence: 99%

“…We resize the images of both datasets to 640x640. The experimental setup is reproduced from [8] except that we average the results over ten trials instead of three.…”

Section: Datasetsmentioning

confidence: 99%

See 2 more Smart Citations

Using filter banks in Convolutional Neural Networks for texture classification

Andrearczyk

Whelan

2016

Pattern Recognition Letters

243

156

View full text Add to dashboard Cite

Deep learning has established many new state of the art solutions in the last decade in areas such as object, scene and speech recognition. In particular Convolutional Neural Network (CNN) is a category of deep learning which obtains excellent results in object detection and recognition tasks. Its architecture is indeed well suited to object analysis by learning and classifying complex (deep) features that represent parts of an object or the object itself. However, some of its features are very similar to texture analysis methods. CNN layers can be thought of as filter banks of complexity increasing with the depth. Filter banks are powerful tools to extract texture features and have been widely used in texture analysis. In this paper we develop a simple network architecture named Texture CNN (T-CNN) which explores this observation. It is built on the idea that the overall shape information extracted by the fully connected layers of a classic CNN is of minor importance in texture analysis. Therefore, we pool an energy measure from the last convolution layer which we connect to a fully connected layer. We show that our approach can improve the performance of a network while greatly reducing the memory usage and computation.

show abstract

Section: Results On Larger Imagesmentioning

confidence: 68%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Using filter banks in Convolutional Neural Networks for texture classification

Andrearczyk

Whelan

2016

Pattern Recognition Letters

243

156

View full text Add to dashboard Cite

show abstract

“…The development of modelling regression to acquire the best estimation of forest carbon storage would be a new direction, especially (Piao et al 2005a, b) after the breakthrough in the field of deep learning. Some convolutional neural network algorithm (CNN) may have potential for estimating forest carbon stocks combine with remote sensing images, such as Alex (Hafemann et al 2014 (Pietsch and Hasenauer 2002). More than 80% of the forests in China belong to immature forests with young and poor quality secondary forests, which were the main reasons for the low carbon storage and carbon density of forests in China (Liu et al 2000;Zhao and Zhou 2004).…”

Section: Research Trendsmentioning

confidence: 99%

Review on carbon storage estimation of forest ecosystem and applications in China

2019

View full text Add to dashboard Cite

Background: The accuracy in estimating forest ecosystem carbon storage has drawn extensive attention of researchers in the field of global climate change. However, incomparable data sources and various estimation methods have led to significant differences in the estimation of forest carbon storage at large scales.Methods: In this study, we reviewed fundamental types of forest carbon storage estimation methods and their applications in China.Results: Results showed that the major forest carbon storage estimation methods were classified into 3 major categories and 15 subcategories focusing on vegetation carbon storage estimation, soil carbon storage estimation, and litter carbon storage estimation, respectively. The application in China showed that there have been 3 development stages of research in China since the 1990s. Studies of forest carbon storage estimation in province scales were conducted more frequently in the northeastern, eastern and southwestern provinces such as Zhejiang, Heilongjiang and Sichuan with high forest coverage or large forest area. Inventory-based methods, soil type method, and biomass model were the main forest estimation methods used in China, focusing on vegetation, soil and litter carbon storage estimation respectively. Total forest carbon storage of China was approximate 28.90 Pg C, and the average vegetation carbon density (42.04 ± 5.39 Mg·ha − 1 ) was much lower than that of the whole world (71.60 Mg·ha − 1 ). Vegetation carbon density from average biomass method was the highest (57.07 Mg·ha − 1 ) through comparing nine types of vegetation carbon storage estimation methods applied during 1989 to 1993. Conclusions:Many studies on forest carbon storages have been carried out in China at patch scales or regional scales. These efforts enabled the research of forest carbon storage to reach a relatively advanced stage. Meanwhile, the accumulation of massive research data provides the basis for subsequent research work. Some challenges are also existing. This review could provide a reference for more accurate estimation of forest carbon storage in the future.

show abstract

“…Kapp et al [17] assessed multiple feature sets using a quadtree-based approach and reported recognition rates of 95 and 88 % for the microscopic [25] and macroscopic [34] databases, respectively. Hafemann et al [13] took a different approach, and instead of using textural descriptors, they used the images to train a convolutional neural network (CNN). Using the datasets proposed in [25] and [34], they reported accuracies of 95 and 97 %, respectively.…”

Section: Introductionmentioning

confidence: 99%

Forest species recognition based on dynamic classifier selection and dissimilarity feature vector representation

Martins

Oliveira

Britto

et al. 2015

Machine Vision and Applications

View full text Add to dashboard Cite

Multiple classifiers on the dissimilarity space are proposed to address the problem of forest species recognition from microscopic images. To that end, classical texturebased features such as Gabor filters, local binary patterns (LBP) and local phase quantization (LPQ), as well as two keypoint-based features, the scale-invariant feature transform (SIFT) and the speeded up robust features (SURF), are used to generate a pool of diverse classifiers on the dissimilarity space. A comprehensive set of experiments on a database composed of 2,240 microscopic images from 112 different forest species was used to evaluate the performance of each individual classifier of the generated pool, the combination of all classifiers, and different dynamic selection of classifiers (DSC) methods. The best result (93.03 %) was observed by incorporating probabilistic information in a DSC method based on multiple classifier behavior.

show abstract

Forest Species Recognition Using Deep Convolutional Neural Networks

Cited by 102 publications

References 12 publications

Using filter banks in Convolutional Neural Networks for texture classification

Using filter banks in Convolutional Neural Networks for texture classification

Review on carbon storage estimation of forest ecosystem and applications in China

Forest species recognition based on dynamic classifier selection and dissimilarity feature vector representation

Contact Info

Product

Resources

About