The UAV technology is gradually maturing and can provide extremely powerful support for smart agriculture and precise monitoring. Currently, there is no dataset related to green walnuts in the field of agricultural computer vision. Thus, in order to promote the algorithm design in the field of agricultural computer vision, we used UAV to collect remote-sensing data from 8 walnut sample plots. Considering that green walnuts are subject to various lighting conditions and occlusion, we constructed a large-scale dataset with a higher-granularity of target features - WalnutData. This dataset contains a total of 30,240 images and 706,208 instances, and there are 4 target categories: being illuminated by frontal light and unoccluded (A1), being backlit and unoccluded (A2), being illuminated by frontal light and occluded (B1), and being backlit and occluded (B2). Subsequently, we evaluated many mainstream algorithms on WalnutData and used these evaluation results as the baseline standard. The dataset and all evaluation results can be obtained at https://github.com/1wuming/WalnutData.

Our paper proposes a direct sparse visual odometry method that combines event and RGB-D data to estimate the pose of agile-legged robots during dynamic locomotion and acrobatic behaviors. Event cameras offer high temporal resolution and dynamic range, which can eliminate the issue of blurred RGB images during fast movements. This unique strength holds a potential for accurate pose estimation of agile-legged robots, which has been a challenging problem to tackle. Our framework leverages the benefits of both RGB-D and event cameras to achieve robust and accurate pose estimation, even during dynamic maneuvers such as jumping and landing a quadruped robot, the Mini-Cheetah. Our major contributions are threefold: Firstly, we introduce an adaptive time surface (ATS) method that addresses the whiteout and blackout issue in conventional time surfaces by formulating pixel-wise decay rates based on scene complexity and motion speed. Secondly, we develop an effective pixel selection method that directly samples from event data and applies sample filtering through ATS, enabling us to pick pixels on distinct features. Lastly, we propose a nonlinear pose optimization formula that simultaneously performs 3D-2D alignment on both RGB-based and event-based maps and images, allowing the algorithm to fully exploit the benefits of both data streams. We extensively evaluate the performance of our framework on both public datasets and our own quadruped robot dataset, demonstrating its effectiveness in accurately estimating the pose of agile robots during dynamic movements.

Fully automatic segmentation methods, such as liver and liver tumor segmentation, brain and brain tumor segmentation, optic disc segmentation, cell segmentation, lung segmentation, pulmonary nodule segmentation, and cardiac image segmentation [2], are essential for the diagnosis of serious diseases [3]. Therefore, it is important to improve the efficiency and accuracy of medical image segmentation methods. Medical image segmentation involves segmenting specific organs (e.g., the pancreas, liver, and bladder), determining certain functional parts of an organ (e.g., cardiac segmentation and retinal vessel segmentation), and identifying tumors in the organs. Medical images can generally be categorized according to the imaging technology and data form. Imaging technology includes X-ray photos, computed tomography, magnetic resonance imaging (MRI), and ultrasound imaging. Raw measurements are transformed into pixelated imaging data as part of the standard process. Although the original data are mostly three-dimensional images, two-dimensional slices are often created according to clinical procedure protocols that target specific applications. Most medical image segmentation methods are designed for two-dimensional slices.

While the basic coding for neural network(NN) is ready, we could try to build a simple hypothesis for multi-class classification, a recognizer for classifying hand-written digits is a good topic for practicing. Let's kick off this demo from "Getting Started Project: Digit Recognizer" on the kaggle community. Our goal is clear: giving a grayscale image with 28x28 pixels, identify the digit the picture shows from '0' to '9'. Concretely, our raw features are the grayscale values at every pixel, with those values, NN attempts to figure out the relationship between pixels, and learning the stroke patterns from different digits, via the supervised learning process, NN could find out the most probable digit the pixels formed. These could be summarized in two facts: (1) the size of the feature space(input layer size) is equal to 784 since all grayscale values may vary independently, (2) we need 10 labels(output layer size) to indicate the digits from '0' to '9'.

You can already control your mobile devices with touch, gesture, or voice commands. Now Finnish researcher Olli Lahdenoja is introducing another mechanism: facial recognition. Lahdenoja, a researcher at the University of Turku, has published a doctoral dissertation which develops a facial recognition method that can fit onto a single electronic chip. The system has the potential to be used in devices such as smartwatches, as well as mobile handsets, thanks to its compact size and low power consumption. "[Using this system] a smartwatch could, for example, switch on when a user looks at it. What's more, the recognition system could be used to identify a person when they access an online service," says Lahdenoja.