A Temporal Boosted YOLO-Based Model for Birds Detection around Wind Farms

Toward Versatile Small Object Detection with Temporal-YOLOv8

Deep learning has become the preferred method for automated object detection, but the accurate detection of small objects remains a challenge due to the lack of distinctive appearance features. Most deep learning-based detectors do not exploit the temporal information that is available in video, even though this context is often essential when the signal-to-noise ratio is low. In addition, model development choices, such as the loss function, are typically designed around medium-sized objects. Moreover, most datasets that are acquired for the development of small object detectors are task-specific and lack diversity, and the smallest objects are often not well annotated. In this study, we address the aforementioned challenges and create a deep learning-based pipeline for versatile small object detection. With an in-house dataset consisting of civilian and military objects, we achieve a substantial improvement in YOLOv8 (baseline mAP = 0.465) by leveraging the temporal context in video and data augmentations specifically tailored to small objects (mAP = 0.839). We also show the benefit of having a carefully curated dataset in comparison with public datasets and find that a model trained on a diverse dataset outperforms environment-specific models. Our findings indicate that small objects can be detected accurately in a wide range of environments while leveraging the speed of the YOLO architecture.

Automatic Detection of Feral Pigeons in Urban Environments Using Deep Learning

The overpopulation of feral pigeons in Hong Kong has significantly disrupted the urban ecosystem, highlighting the urgent need for effective strategies to control their population. In general, control measures should be implemented and re-evaluated periodically following accurate estimations of the feral pigeon population in the concerned regions, which, however, is very difficult in urban environments due to the concealment and mobility of pigeons within complex building structures. With the advances in deep learning, computer vision can be a promising tool for pigeon monitoring and population estimation but has not been well investigated so far. Therefore, we propose an improved deep learning model (Swin-Mask R-CNN with SAHI) for feral pigeon detection. Our model consists of three parts. Firstly, the Swin Transformer network (STN) extracts deep feature information. Secondly, the Feature Pyramid Network (FPN) fuses multi-scale features to learn at different scales. Lastly, the model's three head branches are responsible for classification, best bounding box prediction, and segmentation. During the prediction phase, we utilize a Slicing-Aided Hyper Inference (SAHI) tool to focus on the feature information of small feral pigeon targets. Experiments were conducted on a feral pigeon dataset to evaluate model performance. The results reveal that our model achieves excellent recognition performance for feral pigeons.

Intelligent Detection Method for Wildlife Based on Deep Learning

Wildlife is an important part of natural ecosystems and protecting wildlife plays a crucial role in maintaining ecological balance. The wildlife detection method for images and videos based on deep learning can save a lot of labor costs and is of great significance and value for the monitoring and protection of wildlife. However, the complex and changing outdoor environment often leads to less than satisfactory detection results due to insufficient lighting, mutual occlusion, and blurriness. The TMS-YOLO (Takin, Monkey, and Snow Leopard-You Only Look Once) proposed in this paper is a modification of YOLOv7, specifically optimized for wildlife detection. It uses the designed O-ELAN (Optimized Efficient Layer Aggregation Networks) and O-SPPCSPC (Optimized Spatial Pyramid Pooling Combined with Cross Stage Partial Channel) modules and incorporates the CBAM (Convolutional Block Attention Module) to enhance its suitability for this task. In simple terms, O-ELAN can preserve a portion of the original features through residual structures when extracting image features, resulting in more background and animal features. However, O-ELAN may include more background information in the extracted features. Therefore, we use CBAM after the backbone to suppress background features and enhance animal features. Then, when fusing the features, we use O-SPPCSPC with fewer network layers to avoid overfitting. Comparative experiments were conducted on a self-built dataset and a Turkish wildlife dataset. The results demonstrated that the enhanced TMS-YOLO models outperformed YOLOv7 on both datasets. The mAP (mean Average Precision) of YOLOv7 on the two datasets was 90.5% and 94.6%, respectively. In contrast, the mAP of TMS-YOLO in the two datasets was 93.4% and 95%, respectively. These findings indicate that TMS-YOLO can achieve more accurate wildlife detection compared to YOLOv7.

YOLOv3_ReSAM: A Small-Target Detection Method

Small targets in long-distance aerial photography have the problems of small size and blurry appearance, and traditional object detection algorithms face great challenges in the field of small-object detection. With the collection of massive data in the information age, traditional object detection algorithms have been gradually replaced by deep learning algorithms and have an advantage. In this paper, the YOLOV3-Tiny backbone network is augmented by using the pyramid structure of image features to achieve multi-level feature fusion prediction. In order to eliminate the loss of spatial feature information and hierarchical information caused by pooling operations in convolution processes and multi-scale operations in multi-layer structures, a spatial attention mechanism based on residual structure is proposed. At the same time, the idea of reinforcement learning is introduced to guide bounding box regression on the basis of the rough positioning of the native boundary regression strategy, and the variable IoU calculation method is used as the evaluation index of the reward function, and the boundary regression model based on the reward mechanism is proposed for fine adjustment. The VisDrone2019 data set was selected as the experimental data support. Experimental results show that the mAP value of the improved small-object detection model is 33.15%, which is 11.07% higher than that of the native network model, and the boundary regression accuracy is improved by 23.74%.