YijingYang

May your Christmas sparkle with moments of love, laughter and goodwill. And may the year ahead be full of contentment and joy. Wish all our fellows a Merry Christmas!

The task of Visual Dialogue involves a conversation between an agent system and a human end-user regarding visual information presented. The conversation consists of multiple rounds of questions and answers. The key challenge for the agent system is to answer the questions of human users with meaningful information while keeping the conversation flow contiguous and natural. The current visual dialogue systems can be divided into two tracks, generative models and discriminative models. The discriminative models cannot directly generate a response but choose a response out of a pool of candidate responses. Although the discriminative models have achieved surprising results, they are usually not applicable in real scenarios where no candidate response pool is available. On the other hand, the generative models can directly generate a response based on the input information. However, most generative models based on maximum likelihood estimation (MLE) approach suffer from the tendency of generating generic responses.

We present a novel approach that incorporates a multi-modal recurrently guided attention mechanism with a simple yet effective training scheme to generate high quality responses in the Visual Dialogue interaction model. Our attention mechanism combines attentions globally from multiple modalities (e.g., image, text questions and dialogue history), and refines them locally and simultaneously for each modality. Generators using typical MLE-based methods only learn from good answers, and consequently tend to generate safe or generic responses. The new training scheme with weighted likelihood estimation (WLE) penalizes generic responses for unpaired questions in the training and enables the generator to learn from poor answers as well.

On benchmark dataset, our proposed Visual Dialogue system demonstrates state-of-the-art performance with improvement of 5.81% and 5.28 on recall@10 and mean rank, respectively.

–By Heming Zhang

Humans can accumulate and maintain the knowledge learned from previous tasks and use it seamlessly in learning new tasks and solving new problems — learning new concepts over time is a core characteristic of human learning. Therefore, it is desirable to have a computer vision system that can learn incrementally about new classes when training data for them becomes available, as this is a necessary step towards the ultimate goal of building real intelligent machines that learn like humans.

Despite the recent success of deep learning in computer vision for a broad range of tasks, classical training paradigm of deep models is ill-equipped for incremental learning (IL). Most deep neural networks can only be trained in batch mode in which the complete dataset is given and all classes are known prior to training. However, the real world is dynamic and new categories of interest can emerge over time. Re-training a model from scratch whenever a new class is encountered is prohibitively expensive due to the huge data storage requirements and computational cost.  Directly fine-tuning the existing model on only the data of new classes using stochastic gradient descent (SGD) optimization is not a better approach either, as this might lead to the notorious catastrophic forgetting effect, which refers to the severe performance degradation on old tasks.

Existing IL approaches attempting to overcome catastrophic forgetting tend to produce a model that is biased towards either the old classes or new classes, unless with the help of exemplars of the old data. To address this issue, we propose a class-incremental learning paradigm called Deep Model Consolidation (DMC), which works well even when the original training data is not available. The idea is to train a model on the [...]

Tianhorng Chang, an MCL alumnus, entered the PhD/EE program at USC in Spring 1989, passed his screening exam on December 4, 1989, and became the third PhD student of Dr. Kuo in Spring 1990. Tianhorng took his qualifying exam on May 7, 1992. He was originally scheduled to take his defense in 1993. However, he could not do it but passed away in 1994 because of cancer.

Tianhorng Chang completed 46 units at USC and transferred in another 30 units from his previous MSEE program at the University of Florida. His graduate GPA at USC was 3.93. His seminal paper titled with “Texture Analysis and Classification with Tree-Structured Wavelet Transform” published by IEEE Trans. on Image Processing (1993) received 2000 citations. Due to Tianhorng’s near-complete curriculum and impactful research, USC granted Tianhorng’s posthumous PhD degree in 2018 August.

Tianhorng received his BSEE degree from the National Tsinghua University, Taiwan, in 1983. A ceremony to award the posthumous PhD degree to Tianhorng’s family was held in the National Tsinghua University on November 18, 2018. Many of his undergraduate classmates and MCL alumni in Taiwan attended the award ceremony. During the ceremony, Professor Kuo and several participants shared their memory on Tianhorng. A video clip featuring short speeches given by Professor Richard Leahy (the current Chair of USC/EE), Professor Sandeep Gupta (the former Chair of USC/EE) and Ms. Diane Demetras was also played. Tianhorng’s younger brother received the posthumous PhD degree on behalf of his family. The whole ceremony took about 1 hour.

Professor Kuo said, “Hope this award ceremony brings some comfort to Tianhorng’s family and Tianhorng’s story will encourage a younger generation to pursuit advanced research as long as they have strong passion.”

Image inpainting is the task to reconstruct the missing region in an image with plausible contents based on its surrounding context, which is a common topic of low-level computer vision. Making use of this technique, people could restore damaged images or remove unwanted objects from images or videos. In this task, our goal is to not only fill in the plausible contexts with realistic details but also make the inpainted area coherent with the contexts as well as the boundaries.

In spite of recent progress of deep generative models, generating high-resolution images remains a difficult task. This is mainly because modeling the distribution of pixels is difficult and the trained models easily introduce blurry components and artifacts when the dimensionality becomes high. Following [1] which proposes to synthesize an image based on joint optimization of image context and texture constraints, we divide the task into inference and translation as two separate steps and model each step with a deep neural network [2]. We also use simple heuristics to guide the propagation of local textures from the boundary to the hole. Meanwhile, we introduce a novel block-wise procedural training scheme to stabilize the training and propose a new strategy called adversarial annealing to reduce the artifacts [3].

On the other hand, we observe that existing methods based on generative models don’t exploit the segmentation information to constrain the object shapes, which usually lead to blurry results on the boundary. To tackle this problem, we propose to introduce the semantic segmentation information, which disentangles the inter-class difference and intra-class variation for image inpainting [4]. This leads to much clearer recovered boundary between semantically different regions and better texture within semantically consistent segments. Experiments on multiple public datasets show [...]