The number of trees and their spatial distribution are key information for forest management. In recent years, deep learning-based approaches have been proposed and shown promising results in lowering the expensive labor cost of a forest inventory. In this paper, we propose a new efficient deep learning model called density transformer or DENT for automatic tree counting from aerial images. The architecture of DENT contains a multi-receptive field convolutional neural network to extract visual feature representation from local patches and their wide context, a transformer encoder to transfer contextual information across correlated positions, a density map generator to generate spatial distribution map of trees, and a fast tree counter to estimate the number of trees in each input image. We compare DENT with a variety of state-of-art methods, including one-stage and two-stage, anchor-based and anchor-free deep neural detectors, and different types of fully convolutional regressors for density estimation. The methods are evaluated on a new large dataset we built and an existing cross-site dataset. DENT achieves top accuracy on both datasets, significantly outperforming most of the other methods. We have released our new dataset, called Yosemite Tree Dataset, containing a 10 km² rectangular study area with around 100k trees annotated, as a benchmark for public access.

Computational protein structure prediction is very important for many applications in bioinformatics. In the process of predicting protein structures, it is essential to accurately assess the quality of generated models. Although many single-model quality assessment (QA) methods have been developed, their accuracy is not high enough for most real applications. In this paper, a new approach based on C-a atoms distance matrix and machine learning methods is proposed for single-model QA and the identification of native-like models. Different from existing energy/scoring functions and consensus approaches, this new approach is purely geometry based. Furthermore, a novel algorithm based on deep learning techniques, called DL-Pro, is proposed. For a protein model, DL-Pro uses its distance matrix that contains pairwise distances between two residues' C-a atoms in the model, which sometimes is also called contact map, as an orientation-independent representation. From training examples of distance matrices corresponding to good and bad models, DL-Pro learns a stacked autoencoder network as a classifier. In experiments on selected targets from the Critical Assessment of Structure Prediction (CASP) competition, DL-Pro obtained promising results, outperforming state-of-the-art energy/scoring functions, including OPUS-CA, DOPE, DFIRE, and RW.

Feature extraction is key in understanding and modeling of physiological data. Traditionally hand-crafted features are chosen based on expert knowledge and then used for classification or regression. To determine important features and pick the effective ones to handle a new task may be labor-intensive and time-consuming. Moreover, the manual process does not scale well with new or large-size tasks. In this work, we present a system based on Deep Belief Networks (DBNs) that can automatically extract features from raw physiological data of 4 channels in an unsupervised fashion and then build 3 classifiers to predict the levels of arousal, valance, and liking based on the learned features. The classification accuracies are 60.9%, 51.2%, and 68.4%, respectively, which are comparable with the results obtained by Gaussian Naïve Bayes classifier on the state-of-the-art expert designed features. These results suggest that DBNs can be applied to raw physiological data to effectively learn relevant features and predict emotions.

As the next-generation sequencing (NGS) technologies producing hundreds of millions of reads every day, a tremendous computational challenge is to map NGS reads to a given reference genome efficiently. However, existing methods of all-mappers, which aim at finding all mapping locations of each read, are very time consuming. The majority of existing all-mappers consist of 2 main parts, filtration and verification. This work significantly reduces verification time, which is the dominant part of the running time.

Prediction of protein backbone torsion angles (Psi and Phi) can provide important information for protein structure prediction and sequence alignment. Existing methods for Psi-Phi angle prediction have significant room for improvement. In this paper, a new deep residual inception network architecture, called DeepRIN, is proposed for the prediction of Psi-Phi angles. The input to DeepRIN is a feature matrix representing a composition of physico-chemical properties of amino acids, a 20-dimensional position-specific substitution matrix (PSSM) generated by PSI-BLAST, a 30-dimensional hidden Markov Model sequence profile generated by HHBlits, and predicted eight-state secondary structure features. DeepRIN is designed based on inception networks and residual networks that have performed well on image classification and text recognition. The architecture of DeepRIN enables effective encoding of local and global interatcions between amino acids in a protein sequence to achieve accruacte prediction. Extensive experimental results show that DeepRIN outperformed the best existing tools significantly. Compared to the recently released state-of-the-art tool, SPIDER3, DeepRIN reduced the Psi angle prediction error by more than 5 degrees and the Phi angle prediction error by more than 2 degrees on average. The executable tool of DeepRIN is available for download at http://dslsrv8.cs.missouri.edu/~cf797/MUFoldAngle/.