Camera traps are widely used in wildlife surveys and biodiversity monitoring. Depending on its triggering mechanism, a large number of images or videos are sometimes accumulated. Some literature has proposed the application of deep learning techniques to automatically identify wildlife in camera trap imagery, which can significantly reduce manual work and speed up analysis processes. However, there are few studies validating and comparing the applicability of different models for object detection in real field monitoring scenarios. In this study, we firstly constructed a wildlife image dataset of the Northeast Tiger and Leopard National Park (NTLNP dataset). Furthermore, we evaluated the recognition performance of three currently mainstream object detection architectures and compared the performance of training models on day and night data separately versus together. In this experiment, we selected YOLOv5 series models (anchor-based one-stage), Cascade R-CNN under feature extractor HRNet32 (anchor-based two-stage), and FCOS under feature extractors ResNet50 and ResNet101 (anchor-free one-stage). The experimental results showed that performance of the object detection models of the day-night joint training is satisfying. Specifically, the average result of our models was 0.98 mAP (mean average precision) in the animal image detection and 88% accuracy in the animal video classification. One-stage YOLOv5m achieved the best recognition accuracy. With the help of AI technology, ecologists can extract information from masses of imagery potentially quickly and efficiently, saving much time.

Light field (LF) depth estimation plays a crucial role in many LF-based applications. Existing LF depth estimation methods consider depth estimation as a regression problem, where a pixel-wise L1 loss is employed to supervise the training process. However, the disparity map is only a sub-space projection (i.e., an expectation) of the disparity distribution, which is essential for models to learn. In this paper, we propose a simple yet effective method to learn the sub-pixel disparity distribution by fully utilizing the power of deep networks, especially for LF of narrow baselines. We construct the cost volume at the sub-pixel level to produce a finer disparity distribution and design an uncertainty-aware focal loss to supervise the predicted disparity distribution toward the ground truth. Extensive experimental results demonstrate the effectiveness of our method. Our method significantly outperforms recent state-of-the-art LF depth algorithms on the HCI 4D LF Benchmark in terms of all four accuracy metrics (i.e., BadPix 0.01, BadPix 0.03, BadPix 0.07, and MSE ×100). The code and model of the proposed method are available at https://github.com/chaowentao/SubFocal .

Depth estimation using the Light Field (LF) technique is an essential task with a wide range of practical applications. While mainstream approaches based on multi-view stereo techniques can attain exceptional accuracy by creating finer cost volumes, they are resource-intensive, time-consuming, and often overlook occlusion during cost volume construction. To address these issues and strike a better balance between accuracy and efficiency, we propose an occlusion-aware cascade cost volume for LF depth (disparity) estimation. Our cascaded strategy reduces the sampling number while maintaining a constant sampling interval, enabling the construction of a finer cost volume. We also introduce occlusion maps to enhance accuracy in constructing the occlusion-aware cost volume. Specifically, we first generate a coarse disparity map through a coarse disparity estimation network. Then, we warp the sub-aperture images (SAIs) of adjacent views to the center view based on the coarse disparity map to generate occlusion maps for each SAI by photo-consistency constraints. Finally, we seamlessly incorporate occlusion maps into cascade cost volume to construct an occlusion-aware refined cost volume, allowing the refined disparity estimation network to yield a more precise disparity map. Extensive experiments demonstrate the effectiveness of our method. Compared with the state-of-theart techniques, our method achieves a superior balance between accuracy and efficiency, ranking first in the Q25 metric on the HCI 4D benchmark.

Data augmentation (DA) is an effective approach for enhancing model performance with limited data, such as light field (LF) image super-resolution (SR). LF images inherently possess rich spatial and angular information. Nonetheless, there is a scarcity of DA methodologies explicitly tailored for LF images, and existing works tend to concentrate solely on either the spatial or angular domain. This paper proposes a novel spatial and angular DA strategy named MaskBlur for LF image SR by concurrently addressing spatial and angular aspects. MaskBlur consists of spatial blur and angular dropout two components. Spatial blur is governed by a spatial mask, which controls where pixels are blurred, i.e., pasting pixels between the low-resolution and high-resolution domains. The angular mask is responsible for angular dropout, i.e., selecting which views to perform the spatial blur operation. By doing so, MaskBlur enables the model to treat pixels differently in the spatial and angular domains when super-resolving LF images rather than blindly treating all pixels equally. Extensive experiments demonstrate the efficacy of MaskBlur in significantly enhancing the performance of existing SR methods. We further extend MaskBlur to other LF image tasks such as denoising, deblurring, low-light enhancement, and real-world SR. Code is publicly available at https://github.com/chaowentao/MaskBlur .