Publication

@article{warner2025bark,
  title = {Environmental Sensitivity in AI Tree Bark Detection: Identifying Key Factors for Improving Classification Accuracy},
  author = {Warner, Charles and Wu, Fanyou and Gazo, Rado and Benes, Bedrich and Fei, Songlin},
  journal = {Algorithms},
  year = {2025},
  doi = {10.3390/a18070417},
  volume = {18},
  number = {7}
}

Accurate tree species identification through bark characteristics is essential for effective forest management, but traditionally requires extensive expertise. This study leverages artificial intelligence (AI), specifically the EfficientNet-B3 convolutional neural network, to enhance AI-based tree bark identification, focusing on northern red oak (Quercus rubra), hackberry (Celtis occidentalis), and bitternut hickory (Carya cordiformis) using the CentralBark dataset. We investigated three environmental variables—time of day (lighting conditions), bark moisture content (wet or dry), and cardinal direction of observation—to identify sources of classification inaccuracies. Results revealed that bark moisture significantly reduced accuracy by 8.19% in wet conditions (89.32% dry vs. 81.13% wet). In comparison, the time of day had a significant impact on hackberry (95.56% evening) and northern red oak (80.80% afternoon), with notable chi-squared associations (p < 0.05). Cardinal direction had minimal effect (4.72% variation). Bitternut hickory detection consistently underperformed (26.76%), highlighting morphological challenges. These findings underscore the need for targeted dataset augmentation with wet and afternoon images, alongside preprocessing techniques like illumination normalization, to improve model robustness. Enhanced AI tools will streamline forest inventories, support biodiversity monitoring, and bolster conservation in dynamic forest ecosystems.

@inproceedings{yeo2025phantom,
  title = {PHAnToM: Persona-Based Prompting Has an Effect on Theory-of-Mind Reasoning in Large Language Models},
  author = {Yeo, Gerard and Tan An Ting, Fiona and Jaidka, Kokil and Furniturewala, Shaz and Wu, Fanyou and Xu, Weijie and Jain, Vinija and Chadha, Aman and Liu, Yang and Ng, See-Kiong},
  booktitle = {Proceedings of the International AAAI Conference on Web and Social Media},
  year = {2025},
  doi = {10.1609/icwsm.v19i1.35923}
}

@article{zhang2025falsereject,
  title = {FalseReject: A Resource for Improving Contextual Safety and Mitigating Over-Refusals in LLMs via Structured Reasoning},
  author = {Zhang, Zhehao and Xu, Weijie and Wu, Fanyou and Reddy, Chandan K.},
  journal = {COLM},
  year = {2025},
  arxiv = {2505.08054}
}

@article{warner2024bark,
  title = {CentralBark Image Dataset and Tree Species Classification Using Deep Learning},
  author = {Warner, Charles and Wu, Fanyou and Gazo, Rado and Benes, Bedrich and Fei, Songlin},
  journal = {Algorithms},
  year = {2024}
}

The task of tree species classification through deep learning has been challenging for the forestry community, and the lack of standardized datasets has hindered further progress. Our work presents a solution in the form of a large bark image dataset called CentralBark, which enhances the deep learning-based tree species classification. Additionally, we have laid out an efficient and repeatable data collection protocol to assist future works in an organized manner. The dataset contains images of 25 central hardwood and Appalachian region tree species, with over 19,000 images of varying diameters, light, and moisture conditions. We tested 25 species: elm, oak, American basswood, American beech, American elm, American sycamore, bitternut hickory, black cherry, black locust, black oak, black walnut, eastern cottonwood, hackberry, honey locust, northern red oak, Ohio buckeye, Osage-orange, pignut hickory, sassafras, shagbark hickory silver maple, slippery elm, sugar maple, sweetgum, white ash, white oak, and yellow poplar. Our experiment involved testing three different models to assess the feasibility of species classification using unaltered and uncropped images during the species-classification training process. We achieved an overall accuracy of 83.21% using the EfficientNet-b3 model, which was the best of the three models (EfficientNet-b3, ResNet-50, and MobileNet-V3-small), and an average accuracy of 80.23%.

@article{liu2023can,
  title = {Can language models be used for real-world urban-delivery route optimization?},
  author = {Liu, Yang and Wu, Fanyou and Liu, Zhiyuan and Wang, Kai and Wang, Feiyue and Qu, Xiaobo},
  journal = {The Innovation},
  year = {2023},
  doi = {10.1016/j.xinn.2023.100520},
  publisher = {Elsevier}
}

@article{wu2023ring,
  title = {Data Collection and Deep Learning-Based Detection of Wood Growth Rings},
  author = {Wu, Fanyou and Huang, Yunmei and Benes, Bedrich and Warner, Charles and Gazo, Rado},
  journal = {Information Processing in Agriculture},
  year = {2023}
}

Tree-ring dating enables gathering necessary knowledge about trees, and it is essential in many areas, including forest management and the timber industry. Treering dating can be conducted on either wood's clean cross-sections or tree trunks' rough end cross-sections. However, the measurement process is still time-consuming and frequently requires experts who use special devices, such as stereoscopes. Modern approaches based on image processing using deep learning have been successfully applied in many areas, and they can succeed in recognizing tree rings. While supervised deep learning-based methods often produce excellent results, they also depend on extensive datasets of tediously annotated data. To our knowledge, there are only a few publicly available ring image datasets with annotations. We introduce a new carefully captured dataset of images of hardwood species automatically annotated for tree ring detection. We capture each wood cookie twice, once in the rough form, similar to industrial settings, and then after careful cleaning, that reveals all growth rings. We carefully overlap the images and use them for an automatic ring annotation in the rough data. We then use the Feature Pyramid Network with Resnet encoder that obtains an overall pixel-level area under the curve score of 85.72% and ring level F1 score of 0.7348.

@article{liu2022learning,
  title = {Deep dispatching: A deep reinforcement learning approach for vehicle dispatching on online ride-hailing platform},
  author = {Liu, Yang and Wu, Fanyou and Lyu, Cheng and Li, Shen and Ye, Jiepin and Qu, Xiaobo},
  journal = {Transportation Research Part E},
  year = {2022},
  doi = {10.1016/j.tre.2022.102694}
}

The vehicle dispatching system is one of the most critical problems in online taxi-hailing platforms, which requires adapting the operation and management strategy to the dynamics of demand and supply. In this paper, we propose a single-agent deep reinforcement learning approach for the vehicle repositioning problem by reallocating vacant vehicles to regions with a large demand gap in advance. The simulator and the vehicle repositioning algorithm are designed based on industrial-scale real-world data and the workflow of online taxi-hailing platforms, ensuring the practical value of our approach. Besides, the vehicle repositioning problem is translated in analogy with the load balancing problem in computers. Inspired by the recommendation system, the high concurrency of repositioning requests is addressed by sorting the actions as a recommendation list, whereby matching action with requests. Experiments demonstrate that the proposed approach is superior to the existing ones. It is also worth noting that the proposed approach won first place in the vehicle repositioning task of KDD Cup 2020.

@article{wu2021wood,
  title = {Wood Identification Based on Longitudinal Section Images by Using Deep Learning},
  author = {Wu, Fanyou and Gazo, Rado and Haviarova, Eva and Benes, Bedrich},
  journal = {Wood Science and Technology},
  year = {2021},
  doi = {10.1007/s00226-021-01261-1},
  volume = {55},
  number = {2},
  pages = {553-563}
}

Automatic species identification has the potential to improve the efficacy and automation of wood processing systems significantly. Recent advances in deep learning allowed for the automation of many previously difficult tasks, and in this paper, we investigate the feasibility of using Deep Convolutional Neural Networks (CNNs) for hardwood lumber identification. In particular, we tested two highly effective CNNs (ResNet-50 and DenseNet-121) as well as lightweight MobileNet-V2. Overall, we achieved 98.2% accuracy for 11 common hardwood species classification tasks.

@article{liu2021behavior2vector,
  title = {Behavior2vector: Embedding Users' Personalized Travel Behavior to Vector},
  author = {Liu, Yang and Wu, Fanyou and Lyu, Cheng and Liu, Xin and Liu, Zhiyuan},
  journal = {IEEE Transactions on Intelligent Transportation Systems},
  year = {2021},
  doi = {10.1109/TITS.2021.3078229}
}

We investigate how to effectively and efficiently embed users' personalized travel behaviors to vectors in this paper. Based on an example scenario of travel mode choice in intelligent transportation system, three data structures representing users' travel behaviors are defined, namely heterogeneous graph of users' travel behaviors, user travel behavior k-partite graph, and personalized user travel behavior sentence set. This paper systematically analyzes the principle of existing methods and provides intuitions for the problem of learning travel behavior representation in intelligent transportation system. Then we propose the Behavior2vector, which is an improved method tailored for embedding users' personalized travel behaviors to vectors. In our experiments, we design a travel mode choice model based on machine learning, which uses both hand-crafted basic features and embedded vector features. We further quantify the impact of various factors on travel mode choice and use travel big data to test the hypothesis of traffic assignment models, e.g., travelers always choose the path with the shortest path. In addition, we also compared with the existing graph embedding methods and essentially discussed their advantages and disadvantages.

@inproceedings{wu2021lastmile,
  title = {A hybrid method with rules and optimization to solve the last-mile delivery problems},
  author = {Wu, Fanyou and Liu, Yang},
  booktitle = {Technical Proceedings of the Amazon Last Mile Routing Research Challenge},
  year = {2021},
  editor = {Winkenbach, Matthias and Parks, Steven and Noszek, Joseph}
}

Using historical data to help route planning is significant since the real world is complicated, and the quality of a route is not only defined by its theoretical cost. This report proposed a two-step method that involved learning history and performing classic solutions to vehicle routing problems (VRP). Specifically, the task of the Last Mile Routing Research Challenge is decomposed into two steps: 1) predicting zone level sequence and 2) perform VRP within a single zone. Our method reaches 0.042 locally based on a train test split. The code can be found https://github.com/wufanyou/TLab-Last-Mile.

@inproceedings{Kopp2021,
  title = {Traffic4cast at NeurIPS 2020 - yet more on the unreasonable effectiveness of gridded geo-spatial processes},
  author = {Kopp, Michael and Kreil, David and Neun, Moritz and Jonietz, David and Martin, Henry and Herruzo, Pedro and Gruca, Aleksandra and Soleymani, Ali and Wu, Fanyou and Liu, Yang and Xu, Jingwei and Zhang, Jianjin and Santokhi, Jay and Bojesomo, Alabi and Marzouqi, Hasan Al and Liatsis, Panos and Kwok, Pak Hay and Qi, Qi and Hochreiter, Sepp},
  booktitle = {Proceedings of the NeurIPS 2020 Competition and Demonstration Track},
  year = {2021},
  volume = {133},
  pages = {325--343},
  publisher = {PMLR},
  editor = {Escalante, Hugo Jair and Hofmann, Katja},
  series = {Proceedings of Machine Learning Research},
  month = {06--12 Dec}
}

The IARAI Traffic4cast competition at NeurIPS 2019 showed that neural networks can successfully predict future traffic conditions 15 minutes into the future on simply aggregated GPS probe data in time and space bins, thus interpreting the challenge of forecasting traffic conditions as a movie completion task. U-nets proved to be the winning architecture then, demonstrating an ability to extract relevant features in the complex, real-world, geo-spatial process that is traffic derived from a large data set. The IARAI Traffic4cast challenge at NeurIPS 2020 build on the insights of the previous year and sought to both challenge some assumptions inherent in our 2019 competition design and explore how far this neural network technique can be pushed. We found that the prediction horizon can be extended successfully to 60 minutes into the future, that there is further evidence that traffic depends more on recent dynamics than on the additional static or dynamic location specific data provided and that a reasonable starting point when exploring a general aggregated geo-spatial process in time and space is a U-net architecture.