luoyizhi516's Projects
Out-of-the-box code and models for CMU's object detection and tracking system for multi-camera surveillance videos. Speed optimized Faster-RCNN model. Tensorflow based. Also supports EfficientDet. WACVW'20
[SII2020] Training and validation code for keypoint detection using Stacked-Hourglass networks
Implementation of various methods of single / multi object tracking 🐾🛰
Control vehicle to avoid obstacles using webcam
Optical flow or optic flow is the pattern of apparent motion of objects, surfaces, and edges in a visual scene caused by the relative motion between an observer (an eye or a camera) and the scene. -------------------------- 1. Optical Flow Estimation -------------------------- Implements Lucas-Kanade optical flow estimation, and test it for the two-frame data sets provided viz basketball, grove, and teddy. ------------------- 2. Gaussian Pyramid ------------------- Implements Lucas-Kanade optical flow estimation algorithm in a multi-resolution Gaussian pyramid framework. After experimentally optimizing number of levels for Gaussian pyramid, local window size, and Gaussian width, I have used the same data sets (basketball, grove, and teddy) to find optical flows, visually compared my results with the previous step where I don’t use Gaussian pyramid.
Segmentation of moving objects using dense Optical flow
Code to create shape data set for optical flow tasks
Count the number of people in video stream
A simple and efficient codebase for the optical flow based video object segmentation.
深度学习经典、新论文逐段精读
Code to Detect Pedestrians even in occluded conditions using Faster RCNN with the concept of Repulsion Loss.
This is the implementation code of our paper named Parallel Channel and Position Attention-guided Feature Pyramid for Pig Face Posture Detection (Under Reviewer)
京东猪脸识别比赛
设计并实现了一个基于深度学习、集成学习、迁移学习、GAN等技术的色素性皮肤病自动识别七分类系统。本系统主要由服务端和客户端两个模块组成。服务端基于深度学习、集成学习、迁移学习、GAN等技术实现了对色素性皮肤病自动识别七分类。客户端使用微信小程序和网站(SSM、Springboot)开发。用户通过微信小程序或网站上传图像到服务端,服务端返回所属类别。
An affordable and flexible platform for automated imaging and video recording
a script written in python to generate ply and position map from RGBD
This is basically a clustering algorithm that uses the point cloud data from a 2D lidar to count the number of clusters, the number of points in each cluster and determines the vacant areas through which the robot can move or pass.
using python to process rgbd data and pointcloud data
Implementation of multiple image stitching
PyTorch image models, scripts, pretrained weights -- ResNet, ResNeXT, EfficientNet, EfficientNetV2, NFNet, Vision Transformer, MixNet, MobileNet-V3/V2, RegNet, DPN, CSPNet, and more
PyTorch1.0 tutorials, examples and some books I found 【不定期更新中】整理的PyTorch 1.0 最新版教程、例子和书籍
pytorch model summary, statistic parameters number, memory usage, FLOPs and so on
Implement some models of RGB/RGBD semantic segmentation in PyTorch, easy to run. Such as FCN, RefineNet, PSPNet, RDFNet, 3DGNN, PointNet, DeepLab V3, DeepLab V3 plus, DenseASPP, FastFCN
[内测中]前向式Python环境快捷封装工具,快速将Python打包为EXE并添加CUDA、NoAVX等支持。
Real-time 3D multi-person pose estimation demo on Jetson TX2 with TensorRT.
Directly read RGBD stream (not .bag) from Intel RealSense, save RGB as .jpg and Depth as .Bin.