1 Introduction
2 Materials and Methods
2.1 Dataset Preparation and Preprocessing
2.2 Deep Learning Architecture
2.3 Model Training Process
-
Bounding Box Regression Loss - penalty for wrong anchor box detection, Mean Squared Error calculated based on predicted box location (x, y, h, w);
-
Classification Loss - Cross Entropy calculated for object classification;
-
Objectness Loss - Mean Squared Error calculated for Objectness-Confidence Score (estimation if the anchor box contains an object).
3 Results and Discussion
Class (Species) | Targets | F1 | Precision | Recall | mAP@.5 | mAP@.5:.95 |
---|---|---|---|---|---|---|
All | 3143 | 0.85 | 0.88 | 0.82 | 0.88 | 0.66 |
Wild Boar | 1070 | 0.89 | 0.92 | 0.86 | 0.91 | 0.68 |
Red Deer | 872 | 0.86 | 0.88 | 0.85 | 0.89 | 0.68 |
Red Fox | 356 | 0.94 | 0.93 | 0.94 | 0.97 | 0.75 |
Raccoon Dog | 193 | 0.94 | 0.93 | 0.95 | 0.95 | 0.71 |
European Bison | 176 | 0.81 | 0.89 | 0.76 | 0.85 | 0.68 |
Eurasian Elk | 103 | 0.85 | 0.89 | 0.81 | 0.89 | 0.76 |
Roe Deer | 97 | 0.58 | 0.67 | 0.53 | 0.61 | 0.47 |
Eurasian Red Squirrel | 84 | 0.89 | 0.93 | 0.84 | 0.91 | 0.59 |
Wolf | 71 | 0.87 | 0.89 | 0.85 | 0.91 | 0.73 |
European Badger | 63 | 0.89 | 0.93 | 0.86 | 0.94 | 0.70 |
European Pine Marten | 58 | 0.76 | 0.83 | 0.72 | 0.80 | 0.54 |