FPN (Feature Pyramid Network) has become a basic component of most SoTA one stage object detectors. Many previous studies have repeatedly proved that FPN can caputre better multi-scale feature maps to more precisely describe objects if they are with different sizes. However, for most backbones such VGG, ResNet, or DenseNet, the feature maps at each layer are downsized to their quarters due to the pooling operation or convolutions with stride 2. The gap of downscaling-by-2 is large and makes its FPN not fuse the features smoothly. This paper proposes a new SFPN (Synthetic Fusion Pyramid Network) arichtecture which creates various synthetic layers between layers of the original FPN to enhance the accuracy of light-weight CNN backones to extract objects’ visual features more accurately. Finally, experiments prove the SFPN architecture outperforms either the large backbone VGG16, ResNet50 or light-weight backbones such as MobilenetV2 based on AP score.

The development of lightweight object detectors is essential due to the limited computation resources. To reduce the computation cost, how to generate features plays a significant role. This paper proposes a new lightweight convolution method Cross-Stage Lightweight Module (CSL-M). It combines the Inverted Residual Block (IRB) and Cross-Stage Partial (CSP) concept. Experiments conducted at CIFAR-10 show that the proposed CSL-Net based on CSL-M performs better with fewer FLOPs than the other lightweight backbones. Finally, we use CSL-Net as the backbone to construct a lightweight detector CSL-YOLO, achieving better detection performance with only 43% FLOPs and 52% parameters than Tiny-YOLOv4.

The development of lightweight object detectors is essential due to the limited computation resources. To reduce the computation cost, how to generate redundant features plays a significant role. This paper proposes a new lightweight Convolution method Cross-Stage Lightweight (CSL) Module, to generate redundant features from cheap operations. In the intermediate expansion stage, we replaced Pointwise Convolution with Depthwise Convolution to produce candidate features. The proposed CSL-Module can reduce the computation cost significantly. Experiments conducted at MS-COCO show that the proposed CSL-Module can approximate the fitting ability of Convolution-3x3. Finally, we use the module to construct a lightweight detector CSL-YOLO, achieving better detection performance with only 43% FLOPs and 52% parameters than Tiny-YOLOv4.

Traditional crowd-counting networks suffer from information loss when feature maps are reduced by pooling layers, leading to inaccuracies in counting crowds at a distance. Existing methods often assume correct annotations during training, disregarding the impact of noisy annotations, especially in crowded scenes. Furthermore, using a fixed Gaussian density model does not account for the varying pixel distribution of the camera distance. To overcome these challenges, we propose a Scale-Aware Crowd Counting Network (SACC-Net) that introduces a scale-aware loss function with error-compensation capabilities of noisy annotations. For the first time, we simultaneously model labeling errors (mean) and scale variations (variance) by spatially varying Gaussian distributions to produce fine-grained density maps for crowd counting. Furthermore, the proposed scale-aware Gaussian density model can be dynamically approximated with a low-rank approximation, leading to improved convergence efficiency with comparable accuracy. The lightweight version of our model, named SACC-LW, enhances computational efficiency while preserving accuracy. The superiority and generalization properties of scale-aware loss function are extensively evaluated for different backbone architectures and performance metrics on six public datasets: UCF-QNRF, UCF CC 50, NWPU, ShanghaiTech A, ShanghaiTech B, and JHU. Experimental results also demonstrate that SACC-Net outperforms all state-of-the-art methods, validating its effectiveness in achieving superior crowd-counting accuracy. The source code is available at https://github.com/Naughty725.

Biometrics has become an important research issue in recent years, and the use of deep learning neural networks has made it possible to develop more reliable and efficient recognition systems. Palms have been identified as one of the most promising candidates among various biometrics due to their unique features and easy accessibility. However, traditional palm recognition methods involve 3D point clouds, which can be complex and difficult to work with. To mitigate this challenge, this paper proposes two methods which are Multi-View Projection (MVP) and Light Inverted Residual Block (LIRB).The MVP simulates different angles that observers use to observe palms in reality. It transforms 3D point clouds into multiple 2D images and effectively reduces the loss of mapping 3D data to 2D data. Therefore, the MVP can greatly reduce the complexity of the system. In experiments, MVP demonstrated remarkable performance on various famous models, such as VGG or MobileNetv2, with a particular improvement in the performance of smaller models. To further improve the performance of small models, this paper applies LIRB to build a lightweight 2D CNN called Tiny-MobileNet (TMBNet).The TMBNet has only a few convolutional layers but outperforms the 3D baselines PointNet and PointNet++ in FLOPs and accuracy. The experimental results show that the proposed method can effectively mitigate the challenges of recognizing palms through 3D point clouds of palms. The proposed method not only reduces the complexity of the system but also extends the use of lightweight CNN. These findings have significant implications for developing biometrics and could lead to improvements in various fields, such as access control and security control.

Crop monitoring focuses on detecting and identifying numerous crops within a limited region. A major challenge arises from the fact that the target crops are typically smaller in size compared to the image resolution, as seen in the case of rice plants. For instance, a rice plant may only span a few dozen pixels in an aerial image that comprises thousands to millions of pixels. This size discrepancy hinders the performance of standard detection methods. To overcome this challenge, our proposed solution includes a testing time grid cropping method to reduce the scale gap between rice plants and aerial images, a multi-scale prediction method for improved detection using cropped images based on varying scales, and a mean-NMS to prevent the potential exclusion of promising detected objects during the NMS stage. Furthermore, we introduce an efficient object detector, the Enhanced CSL-YOLO, to expedite the detection process. In a comparative analysis with two advanced models based on the public test set of the AI CUP 2021, our method demonstrated superior performance, achieving notable 4.6% and 2.2% increases in F1 score, showcasing impressive results.