We discuss how a deep learning model may tackle the one-shot learning scenario on image translation. We propose a two-step training strategy to solve the blurry image result due to the lack of training data. Furthermore, we successfully get the grouping information when extracting features. Finally, we show that most people prefer the synthesized images from our model.

Implement Iterative Closest Point Algorithm to matching the golden TFT panel image and testing images to find defect position and classify the defect type.

Design a feature extraction working flow with a neural network to detect new circuit pattern by comparing the similarity between the testing and known circuit patterns.

We present a cascaded conditional adversarial network to construct the transformation between Chinese characters of two different fonts that exhibit a large structural difference. The end-to-end trained model takes only a subset of Chinese characters as training data, and learns to generalize and achieve elaborated handwriting synthesis. The mechanism of the proposed Cascaded-GAN decouples and simplifies the training task. The experimental results show that Cascaded-GAN can imitate the original handwriting styles more naturally, and the rendered handwriting passages look plausible with subtle variations.