Advanced Neural Network Applications

Advancing Wildlife Image Analysis: A Graph Attention Contrastive Learning Approach for Region-Specific Mammal Classification

This paper develops a region-specific AI model for mammal classification using graph attention contrastive learning, achieving high accuracy in recognizing local species and improving data cleaning efficiency.

Object-zoomed training of convolutional neural networks inspired by toddler development improves shape bias

The paper introduces object-zoomed training for CNNs to improve shape bias, aligning CNN performance with human-like object recognition strategies.

Deep Blueprint: A Literature Review and Guide to Automated Image Classification for Ecologists

This paper provides a guide for ecologists to use deep learning for image classification, demonstrating a practical approach with open-source data and models.

InsectMorphoAI: A Deep Learning-Based Software for Automated Estimation of Insect Length, Volume, and Biomass

InsectMorphoAI automates insect length, volume, and biomass estimation from 2D images using two methods, achieving high accuracy and enabling efficient ecological research.

Paying Attention to Other Animal Detections Improves Camera Trap Classification Models

This paper improves camera trap classification models by incorporating self-attention to leverage information from multiple animal detections, enhancing species classification accuracy in both mono- and multi-species scenarios.

Insect Size Matters: Using Image and Dimensions Together Improves Image Classification

This paper improves insect classification by integrating body size information into deep learning models, showing it enhances accuracy, especially in low-data and taxonomic misclassification scenarios.

A General Method for Detection and Segmentation of Terrestrial Arthropods in Images

flatbug, a YOLOv8 instance-segmentation model with multi-scale tile-pyramid inference and polygon-based deduplication, segments terrestrial arthropods across 23 image subdatasets and transfers to held-out subdatasets with a modest F1 penalty.

Automated Specimen Triage for Dark Taxa: Deep Learning Enables Orientation, Sex Identification, and Anatomical Segmentation from Robotic Imaging

This paper uses deep learning to automate specimen triage by identifying sex, orientation, and anatomical regions from robotic imaging of Phoridae specimens.

A study on edge devices for image classification of the Tasmanian devil (Sarcophilus harrisii) for vaccine delivery

This paper evaluates edge devices for image classification of Tasmanian devils to optimize vaccine delivery systems.

Reducing Manual Labour in Forensic Microtrace Recognition with Deep Learning

This paper develops a deep learning model to automate microtrace recognition in forensic microscopy, reducing manual annotation needs by 65% through combined pretraining strategies.

Utilizing CNNs for classification and uncertainty quantification for 15 families of European fly pollinators

This study uses CNNs to classify 15 European fly families and quantify uncertainty, achieving 95.61% accuracy with EfficientNetB4.

Potentials and limitations in the application of Convolutional Neural Networks for mosquito species identification using wing images

This study develops a CNN-based system for mosquito species identification using wing images, achieving high accuracy and demonstrating practical usability in real-world conditions.

Addressing the open world: detecting and segmenting pollen on palynological slides with deep learning

Fusing a small-grain expert detector, fine-tuning across microscopes, human-in-the-loop continual learning and zero-shot SAM-2 segmentation improve detection and isolation of fossil pollen on scanned slides.

A deep learning approach to detect and visualise sexual dimorphism in monomorphic species

This paper uses deep learning to detect sexual dimorphism in sociable weavers, finding higher accuracy than humans and identifying beak regions as key for sex prediction.

The MacqD Deep Learning-based Model for Automatic Detection of Socially Housed Laboratory Macaques

This paper introduces MacqD, a deep learning model for detecting macaques in complex lab environments, achieving high F1-scores across various scenarios.

AInsectID Version 1.1: an Insect Species Identification Software Based on the Transfer Learning of Deep Convolutional Neural Networks

AInsectID Version 1.1 is an insect species identification software using transfer learning with CNNs, achieving 99.65% validation accuracy.

Classification of dog breeds using convolutional neural network models and support vector machine

This study improves dog breed classification accuracy using a combination of CNN features, PCA, GWO, and SVM, achieving 95.24% and 99.34% accuracy for 120 and 76 breeds respectively.

Teaching deep networks to see shape: Lessons from a simplified visual world.

The paper investigates why deep networks rely on color and texture over shape for classification, showing that training methods bias networks toward simpler features.

Towards a Taxonomy Machine. A Training Set of 5.6 Million Arthropod Images

This paper releases a dataset of 5.6 million arthropod images for training machine learning models in taxonomic identification.

Data-centric AI approach for automated wildflower monitoring

The paper introduces a dataset and a PyTorch model for automated wildflower monitoring, achieving a high mAP score for species identification and counting.

SnakeFace: a transfer learning based app for snake classification

The paper demonstrates the use of transfer learning for snake classification, showing high accuracy in species identification using deep learning models.

Improved prediction of behavioral and neural similarity spaces using pruned DNNs

This paper improves prediction of behavioral and neural similarity spaces by pruning DNN embeddings to align with human judgments and neurobiological data.

Detection and measurement of butterfly eyespot and spot patterns using convolutional neural networks

The paper develops two CNNs to detect and measure butterfly spots and eyespots in wing images, achieving high accuracy in distinguishing these patterns from other color patches.

The role of capacity constraints in Convolutional Neural Networks for learning random versus natural data

This paper investigates how capacity constraints in CNNs affect learning of random versus natural data, finding that internal noise most effectively reduces performance on random data to chance levels.

Contextual associations represented both in neural networks and human behavior

The paper shows that contextually related objects are more similar in CNN representations than unrelated ones, and this similarity correlates with human behavior in later network layers.

DeepImageTranslator: a free, user-friendly graphical interface for image translation using deep-learning and its applications in 3D CT image analysis

The paper introduces DeepImageTranslator, a graphical tool for image translation using deep learning, and demonstrates its effectiveness in CT image segmentation and noise reduction.

Biological convolutions improve DNN robustness to noise and generalisation

This paper shows that adding biological filter banks improves DNN robustness to noise and generalisation to out-of-domain images.

Hierarchical sparse coding of objects in deep convolutional neural networks

This study shows that deep convolutional neural networks use a hierarchical sparse coding scheme similar to the primate brain, with increasing sparseness and correlation with object recognition performance.

Depth in convolutional neural networks solves scene segmentation

This paper shows that deeper convolutional neural networks better distinguish objects from backgrounds in scenes, suggesting that scene segmentation can emerge from network depth without explicit mechanisms.

Recurrent convolutional neural networks: a better model of biological object recognition

This paper compares recurrent convolutional neural networks with feedforward models for object recognition, finding recurrent networks more accurate and robust, especially under occlusion and noise.