Detecting moving objects in a video sequence is an important problem in many vision-based applications. In particular, detecting moving objects when the camera is moving is a difficult problem. In this study, we propose a symmetric method for detecting moving objects in the presence of a dynamic background. First, a background compensation method is used to detect the proposed region of motion. Next, in order to accurately locate the moving objects, we propose a convolutional neural network-based method called YOLOv3-SOD for detecting all objects in the image, which is lightweight and specifically designed for small objects. Finally, the moving objects are determined by fusing the results obtained by motion detection and object detection. Missed detections are recalled according to the temporal and spatial information in adjacent frames. A dataset is not currently available specifically for moving object detection and recognition, and thus, we have released the MDR105 dataset comprising three classes with 105 videos. Our experiments demonstrated that the proposed algorithm can accurately detect moving objects in various scenarios with good overall performance.

Murine behavior recognition is widely used in biology, neuroscience, pharmacology, and other aspects of research, and provides a basis for judging the psychological and physiological state of mice. To solve the problem whereby traditional behavior recognition methods only model behavioral changes in mice over time or space, we propose a symmetrical algorithm that can capture spatiotemporal information based on behavioral changes. The algorithm first uses the improved DeepLabCut keypoint detection algorithm to locate the nose, left ear, right ear, and tail root of the mouse, and then uses the ConvLSTM network to extract spatiotemporal information from the keypoint feature map sequence to classify five behaviors of mice: walking straight, resting, grooming, standing upright, and turning. We developed a murine keypoint detection and behavior recognition dataset, and experiments showed that the method achieved a percentage of correct keypoints (PCK) of $87\pm 1$% at three scales and against four backgrounds, while the classification accuracy for the five kinds of behaviors reached $93\pm 1$%. The proposed method is thus accurate for keypoint detection and behavior recognition, and is a useful tool for murine motion behavior recognition.

In this work, we propose a symmetry approach and design a convolutional neural network for mouse pose estimation under scale variation. The backbone adopts the UNet structure, uses the residual network to extract features, and adds the ASPP module into the appropriate residual units to expand the perceptual field, and uses the deep and shallow feature fusion to fuse and process the features at multiple scales to capture the various spatial relationships related to body parts to improve the recognition accuracy of the model. Finally, a set of prediction results based on heat map and coordinate offset is generated. We used our own built mouse dataset and obtained state-of-the-art results on the dataset.

Accurate target detection in natural environments is an important function of the visual systems of vertebrates and has a direct impact on animal survival and environmental adaptation. Existing studies have shown that the mammalian prefrontal cortex plays an important role in target detection. However, target detection mechanisms in brain regions similar to other species, such as the avian nidopallium caudolaterale, have not been well studied. Here, we selected pigeons, known for their excellent target detection ability, as an animal model and studied the dynamic changes in the nidopallium caudolaterale neural network features while they performed a target detection task in a maze. The results showed that the average node degree increased significantly during the target detection process while modularity decreased significantly. This indicated that functional connectivity in pigeon brains was enhanced during the task execution, the frequency of brain interactions increased, and the neural network shifted from distributed processing to more efficient integrated processing. The decoding results based on the average node degree and modularity and the combination of both showed that the accuracy of target decoding corresponding to the combination of both was higher. Taken together, our results confirmed the important role of the above properties for encoding target information. We provided evidence to support the view that the NCL is critical for target detection tasks and that studying key features of its neural network provides a powerful tool for revealing the functional state of the brain.

Numerosity perception is an evolutionarily conserved ability observed across diverse taxa, from insects and fish to birds and primates. Unlike traditional visual categories, numerosity possesses an intrinsic metric structure, making it well suited for quantitatively investigating how categorical representations emerge along the visual hierarchy. While numerical representations are well characterized in high-level associative areas of non-human primates and avian species, neuronal processing of numerosity in upstream regions remains largely unexplored. To address this gap, we recorded single-unit activity in the pigeon entopallium during a delayed match-to-numerosity task. We found a subset of neurons that encoded numerosity independently of the non-numerical feature controlled in the corresponding recording session, exhibiting quasi-monotonic increasing (QMI) or decreasing (QMD) response profiles. Compared with QMI neurons, QMD neurons exhibited a markedly later coding window and stronger category discriminability. Moreover, in both neuronal populations, the normalized response functions and category discriminability were better described by compressed numerosity scales than by a linear scale, with the logarithmic scale showing a modest overall advantage. These findings help refine current theoretical frameworks for how numerosity information is extracted and represented along the avian visual processing hierarchy, providing a comparative basis for cross-species research.

Understanding how birds perceive and recognize objects remains a fundamental question in neuroscience. The entopallium, a key node in the avian tectofugal pathway, is involved in complex visual processing, yet its functional organization remains incompletely understood. Here, we systematically mapped entopallium neuronal responses in pigeons using controlled visual stimuli varying in color, shape, and motion. Entopallium neurons exhibited strong hue selectivity invariant to luminance, pronounced orientation tuning to shape stimuli, and robust direction selectivity to motion. Spatial analyses revealed distinct functional segregation, with color-selective neurons localized anteroventrally, shape-selective neurons dorsally, and motionselective neurons posteriorly. Despite this segregation, a subset of neurons exhibited joint tuning across stimulus dimensions, suggesting an integrated yet modular organizational scheme. Notably, entopallium neurons exhibited a moderate level of feature integration, sharing key characteristics with early to intermediate stages of mammalian visual processing. Together, these findings establish the entopallium as a central hub for multidimensional visual processing in birds and provide new insights into the convergent evolution of sophisticated visual systems across vertebrates.