智能驾驶时代即将来临！随着汽车的大范围普及，辅助驾驶系统的需求越来越大。我们计划利用FPGA快速并行的运算能力，开发一套实时的环视显示辅助驾驶图像处理系统，使得驾驶更加安全舒适。我们将在采用四路鱼眼摄像头作为视频输入，在DE10-NANO平台上经过软硬件算法协同处理，输出俯瞰视角的车辆环视影像。这一套系统将能大大降低驾驶事故发生的概率，大幅降低泊车难度。

PipeCNN是我们团队提出的一种基于OpenCL的深度卷积神经网络FPGA加速器，可以在各种FPGA平台上实现。PipeCNN项目可以公开访问，网址为：https://github.com/doonny/PipeCNN
本次利用我们团队设计的PipeCNN:通用的深度卷积神经网络FPGA加速器，对以下四个应用设计进行演示：
（1）ImageNet图像分类：基于AlexNet网络对ImageNet数据集进行实时分类
（2）基于摄像头的物体识别：通过摄像头采集目标物体图像，并对其进行识别
（3）人脸识别：基于VGG-16网络对给定人脸进行实时的识别
（4）目标检测：基于Faster RCNN网络在分类图像的同时把物体用矩形框圈出来。
PipeCNN is an efficient FPGA accelerator proposed by our team that
can be implemented on a variety of FPGA platforms with reconfigurable performance and cost. The PipeCNN project proposed by our team is openly accessible,you can get it on our github website: https://github.com/doonny/PipeCNN.
We use the PipeCNN- an efficient FPGA accelerator to demonstrate the following four application designs:
(1) ImageNet classification. ImageNet database was used and a number of five hundred pictures were processed on the test board. For AlexNet, the achieved classification speed is 110 ms per imagewe.
(2) Object recognition via camera. we use a USB camera as a video input, interactively intercept a picture from the video, then parallel computing it on the DE10-NANO platform, the final real-time classification of the target object and display in the VNC interface.
(3) Face recognition. In order to prove that our acceleration system is a universal model, in this section we use the VGG-Net network for face recognition experiments.
(4) Object Detection. Finally, we apply our proposed accelerator in the object detection. We use Faster RCNN Net to detect the target and draw it out.

To help machines learn what we human beings are doing via a camera is important. Once it comes true, machines can make different responses to all kinds of human's postures. But the process is very difficult as well, because usually it is very slow and power-consuming, and requires a very large memory space. Here we focus on real-time posture recognition, and try to make the machine "know" what posture we make. The posture recognition system is consisted of DE10-Nano SoC FPGA Kit, a camera, and an HDMI monitor. SoC FPGA captures video streams from the camera, recognizes human postures with a CNN model, and finally shows the original video and classification result (standing, walking, waving, etc.) via HDMI interface.

This proposal presents the implementation of the communication between deaf-mutes and the normal with a hardware design named sign language interaction device adopting SVM and HMM. On the one hand, gesture recognition system is to create a system which understands human gesture and translates them into texts and audio. On the other hand, our oral language can be interpreted into texts displayed on LCD. Ultimately, the system is designed to identify 20 Chinese sign language and also real time hand gesture signs. The proposed system is very convenient and high-efficiency due to FPGA implementation which is highly suitable for control of equipment by the handicapped people.

Control the drone with no constructions but mirroring aerial model.

本系统利用Intel的DE10-nano平台设计并制作了一款自动烟叶识别、分级系统，用以解决我国的烟叶分级工作主要由人工手动完成，存在着分级正确率低、效率低下、主观因素大、成本高等缺点。该系统能够利用机械臂自动拾取烟叶，具有图像采集、模式识别、LAB算法分级、分度盘自动分类、远程显示与控制等功能。作品首先利用运行Linux的ARM硬核驱动USB摄像头，将烟叶的图像数据采集到DDR3的特定物理地址，再通过h2f总线将图像数据传到FPGA的图像处理IP核，该IP核为自定义IP核，能够完成图像格式转换、模式识别及机器学习等功能，作品通过OpenCL完成这些算法的设计，具有实时性强、处理速度快等优点。随后，OpenCL输出的图像处理结果通过f2h总线传送给DDR3，SOC中的ARM读取这些数据，再利用C程序对数据做进一步的分析，实现烟叶的特征提取，得到分级结果。最后，ARM根据烟叶特征及分级信息，指挥传送带以及分度盘的电机完成适当的动作，将不同等级的烟叶放在不同的分度盘中，从而实现烟叶的自动分级。

The traditional supervised learning algorithm of ANN, such as Back-Propagation (BP) algorithm based on gradient descent, is not suitable for SNNs because the information is propagated by a sequence of spikes which are not continuous and differentiable. The STDP learning algorithm represents a Hebbian form of plasticity that adjusts the strength of connections based on the relative timing of a particular neuron's output and input spikes. It simplifies the timing information of spikes by using discrete time models, such as the Spike Response Model (SRM) and Integrate-and-fire (I&F).

本设计基于FPGA实现了一种声源目标定位与识别系统。系统通过麦克风阵列感知环境声场，随后采用数字背景噪声抑制进一步提高信号的信噪比从而得到低噪声、宽动态范围的声场数据，为了补偿环境因素对测量的影响，系统还通过传感器测量当前环境下的温度和气压，用来调整用于计算的声速，最后采用麦克风阵列定位原理实现声源目标的空间定位，并借助特征提取和模式匹配方法进行声源目标识别。

Vision is a gift. However, there are some things human eyes cannot see directly. Therefore machine vision technology developed to help human life. Along with this project we propose solution for multi-purpose application using machine vision technology. We are developing a contactless real-time system to measure human soft-biometric and vital signs using DE10-Nano and D8M camera as main hardware. Moreover, we are providing interfaces and peripheral which able to fulfill target application needs. We believe our system has promising benefits for human life with offers high specification and low-cost technology.

The system mainly USES the camera to collect the image of the intersection, and then detects the traffic and pedestrian situation of the road junction with ultrasonic, infrared and other sensors. The above data is processed by FPGA, and the current traffic light scheme is determined. To achieve the efficiency of vehicles through intersections. It can slow the contradiction between the rapid growth of vehicles and the slow expansion of urban roads, and ensure the safety while reducing the congestion and waiting time for car owners. At the same time, it can achieve the goal of energy conservation and emission reduction, and indirectly contribute to environmental protection.
Because the system is always data processing and control, and not the sequential system, it can give full play to the advantages of FPGA high-speed processing and extremely low delay. Meanwhile, the stability of FPGA also provides an important guarantee for the safe operation of the system.
It is proposed to build an experimental model to test the various functions of the system, and to promote it systematically. Apply for a practical patent.

人工智能与眼科结合产生的疾病诊断平台，通过智能阅片，准确、及时地得出诊断结果，将有效协助医生的诊断，提高诊断效率。目前卷积神经网络广泛应用在图像识别领域，而FPGA具有并行计算和低功耗的特点，因此被广泛应用于卷积神经网络的硬件加速。我们采用摄像头采集眼部图片，在DE10-Nano平台上利用卷积神经网络进行计算，最终得到眼科疾病的诊断结果。另外，未来手持眼底检查设备的应用将大大加快该系统的使用。

如何把壽命終止的垃圾物回收再利用，是個很大的議題，澳洲新南威爾斯大學教授薩哈吉瓦拉，致力於將廢棄的垃圾，變成可回收再造的資源。隨著人們生活品質的提升，使用塑膠製品已經相當的習慣了，但也造成海洋生態的塑膠汙染日益嚴重。近來香港也有抗爭活動希望政府能禁用塑膠製品。將垃圾分為不可燃與可燃等物品分類，已經是大家日常生活中一直在做的事。但是還是有些人不願意配合做垃圾分類，會在垃圾焚化處理時產生空氣污然，會產生致癌物傷害人類的健康。
本系統希望藉由深度影像讓機器能自動判斷出目標物體與所在位置，並由機器手臂正確抓取目標垃圾正確的回收箱。達到以2D顏色之影像與深度影像之3D視覺使用深度學習進行物辨視，進而進行垃圾分類，降低環境污染。