ML expert for the Innovative Ads Creative team

As an rsde at content intellegence team，my main responsibility is to design and apply AI algorithms （CV, NLP) to understand video contents and users.

Project: Learning Distributed Data using Generative Adversarial Networks• Compared the performance 2 GAN models on multiple datasets.• Tuned models on server nodes to learn the distribution of local dataset.• Exploring semi-supervised classification for GAN-generated data.

Project: Offline handwriting recognition with Recurrent Neural Networks• Modified the 2D LSTM network provided by RNNLIB for handwritten Arabic recognition.• Trained an 1D LSTM model using TensorFlow.• Explored the application on Chinese character detection and recognition.Project: Neural network-based Chinese character recognition for camera-captured document images• Trained different CNN models for character recognition.• Compiled the CNN recognition module to Android platform using Android NDK.• Built the Android App.

Project: Semantic segmentation for Advanced Driver Assistance System • Took part in the design of the real-time semantic segmentation model for ADAS. • Designed and implemented a multi-dimensional LSTM layer for current model to extract long-range contextual information from image. • Working on the improvement of LSTM layer

Project: Automatic text detection and recognition system for News videos• Designed a multi-stage scheme for text detection.• Trained a CNN network using Caffe for text and non-text candidates classification.• Verified the method on both Chinese and English datasets.• Published a conference paper on Procedia Computer Science, 2016.

Project: Palm print identification system on Android platform• Designed the palm identification system on Android platform.• Proposed a new palm landmark detection approach for mobile-captured images.• Published a conference paper on Electronic Imaging, 2016.