Scholarlink

<p>This research introduces a new method using deep learning to automatically identify animal footprints (spoors) in images. This can help reduce conflicts between humans and wildlife. The method combines two techniques decoder convolutional neural networks (de-CNNs) and Transformers: de-CNNs, which excel at extracting image features, and Transformers, which are great at recognizing shapes and edges. Our method leverages the strengths Canny edge detection (CannyED) in de-CNN-Transformers combination. This improves both accuracy and understanding of the model’s decisions. Detected edges are segmented and reinforced for robustness. To ensure accurate classification, the system’s learned active contour prioritizes key footprint areas by meticulously examining four critical regions: tarsal pad, spoor width, and hind pad to classify carnivore species. This targeted approach offers advantages: Sharper Vision: CannyED with Transformers improves accuracy in identifying spoors. Clearer Understanding: Focusing on specific features helps us understand how the model makes predictions. Targeted Analysis: Analysing crucial footprint regions increases efficiency and reduces reliance on irrelevant information. Experiments showed this method (de-CNN-Transformer) outperforms existing ones, achieving 95% accuracy in edge detection and 88% in classifying carnivore spoors from 2D aerial imagery. This research presents a promising tool for mitigating human-wildlife conflict through automatic footprint recognition.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Learned active contours via transformer-based deep convolutional neural network using canny edge detection algorithm

  • 期刊论文

  • 发表日期：

    2025年1月17日

查看全文
  • Johnas Omanwa Maranga,
  • Justine John Nnko,
  • Shengwu Xiong

摘要

This research introduces a new method using deep learning to automatically identify animal footprints (spoors) in images. This can help reduce conflicts between humans and wildlife. The method combines two techniques decoder convolutional neural networks (de-CNNs) and Transformers: de-CNNs, which excel at extracting image features, and Transformers, which are great at recognizing shapes and edges. Our method leverages the strengths Canny edge detection (CannyED) in de-CNN-Transformers combination. This improves both accuracy and understanding of the model’s decisions. Detected edges are segmented and reinforced for robustness. To ensure accurate classification, the system’s learned active contour prioritizes key footprint areas by meticulously examining four critical regions: tarsal pad, spoor width, and hind pad to classify carnivore species. This targeted approach offers advantages: Sharper Vision: CannyED with Transformers improves accuracy in identifying spoors. Clearer Understanding: Focusing on specific features helps us understand how the model makes predictions. Targeted Analysis: Analysing crucial footprint regions increases efficiency and reduces reliance on irrelevant information. Experiments showed this method (de-CNN-Transformer) outperforms existing ones, achieving 95% accuracy in edge detection and 88% in classifying carnivore spoors from 2D aerial imagery. This research presents a promising tool for mitigating human-wildlife conflict through automatic footprint recognition.