1. Built a web application based on neural networks for flooding control and pollution monitoring of the project named Visual Inspection, implemented in Flask, Bootstrap and Caffe(Python).

2. Innovated a fully convolutional neural network for the segmentation of reservoir surveillance video, accomplished 85.2% Pixel Accuracy and 31.2% Mean IoU.

3. Developed an image annotation tool named ‘inSpecLabel’ with PyQt5 to label objects with pixel accuracy.

4. Applied HOG and SVM-based feature extraction and Hough-transform-based line recognition to measure the readings of the detected water gauges.

5. Developed a fine-tuning strategy to train an industrial inspection models based on Faster-R-CNN for detecting the Printed Circuit Board defects among 25 classes, got a mAP value among all classes above 85% and improved the detecting accuracy by more than 10% approaching 95% comparing to the traditional AOI Inspection System (83%).

1. Used MVC framework to develop a website to popularize the Taobao Design Guidelines to customers and traders.

2. Designed and developed an HTML5 automatic generator “Duang”, used Node.js to make project lightweight, solved asynchronous and callback when design complex interaction effects.

3. Developed a web Mahjong game with Cocos2d-JS game engine, integrated into the promotion pages of ‘Taobao Double Twelve’, a Chinese version of ‘Black Friday’, served 10 million users.