Machine Learning ● Pixsee Baby Camera a. Face detection and attributes(age, gender, etc.) model training for the "Baby Bubble Tree" service. b. Object detection model training and image-based product recommendations. c. People search with unsupervised clustering. d. Model Interpretability and embedding visualization(t-SNE). e. Design and implement the process of model benchmarking, data statistics, and error analysis.

Machine Learning ● Lollipop Baby camera a. Baby detection model training. b. Use GAN, OpenCV, and crawler to generate and collect data. c. Use the Guided Grad-CAM to show models looking for CNN visualization. ● Face Attribute a. Face detection and face attributes multi-task model training. ● Mobile Library a. Build Android MACE library for SSD, MTCNN, segmentation, and classification. b. Build iOS Core ML library for SSD, vision face, and classification.

Machine Learning ● Pose Estimation – Use OpenPose to perform multiple people estimation. a. Customed training OpenPose to perform multiple people keypoint detection. b. Build real-time and offline Android tflite demo application. Back-End ● Use the Azure service to build the Android Launcher application back-end. a. Request news regularly and insert parsed data into the SQL dataset via Mobile Apps Jobs. b. Created thumbnail via VM and uploaded images regularly to blob storage. Web ● Build age and gender web via the Bootstrap framework. Android Application ● “Calendar” widget and year view features.