One-shot weight sharing methods have recently drawn great attention in neural architecture search due to high efficiency and competitive performance. However, weight sharing across models has an inherent deficiency, i.e., insufficient training of subnetworks in the hypernetwork. To alleviate this problem, we present a simple yet effective architecture distillation method. The central idea is that subnetworks can learn collaboratively and teach each other throughout the training process, aiming to boost the convergence of individual models. We introduce the concept of prioritized path, which refers to the architecture candidates exhibiting superior performance during training. Distilling knowledge from the prioritized paths is able to boost the training of subnetworks. Since the prioritized paths are changed on the fly depending on their performance and complexity, the final obtained paths are the cream of the crop. We directly select the most promising one from the prioritized paths as the final architecture, without using other complex search methods, such as reinforcement learning or evolution algorithms. The experiments on ImageNet verify such path distillation method can improve the convergence ratio and performance of the hypernetwork, as well as boosting the training of subnetworks. The discovered architectures achieve superior performance compared to the recent MobileNetV3 and EfficientNet families under aligned settings. Moreover, the experiments on object detection and more challenging search space show the generality and robustness of the proposed method. Code and models are available at https://github.com/microsoft/cream.git.
Most of the current action localization methods follow an anchor-based pipeline: depicting action instances by pre-defined anchors, learning to select the anchors closest to the ground truth, and predicting the confidence of anchors with refinements. Pre-defined anchors set prior about the location and duration for action instances, which facilitates the localization for common action instances but limits the flexibility for tackling action instances with drastic varieties, especially for extremely short or extremely long ones. To address this problem, this paper proposes a novel anchor-free action localization module that assists action localization by temporal points. Specifically, this module represents an action instance as a point with its distances to the starting boundary and ending boundary, alleviating the pre-defined anchor restrictions in terms of action localization and duration. The proposed anchor-free module is capable of predicting the action instances whose duration is either extremely short or extremely long. By combining the proposed anchor-free module with a conventional anchor-based module, we propose a novel action localization framework, called A2Net. The cooperation between anchor-free and anchor-based modules achieves superior performance to the state-of-the-art on THUMOS14 (45.5% vs. 42.8%). Furthermore, comprehensive experiments demonstrate the complementarity between the anchor-free and the anchor-based module, making A2Net simple but effective.
High-quality video inpainting that completes missing regions in video frames is a promising yet challenging task. State-of-the-art approaches adopt attention models to complete a frame by searching missing contents from reference frames, and further complete whole videos frame by frame. However, these approaches can suffer from inconsistent attention results along spatial and temporal dimensions, which often leads to blurriness and temporal artifacts in videos. In this paper, we propose to learn a joint Spatial-Temporal Transformer Network (STTN) for video inpainting. Specifically, we simultaneously fill missing regions in all input frames by self-attention, and propose to optimize STTN by a spatial-temporal adversarial loss. To show the superiority of the proposed model, we conduct both quantitative and qualitative evaluations by using standard stationary masks and more realistic moving object masks. Demo videos are available at https://github.com/researchmm/STTN.
Anchor-based Siamese trackers have achieved remarkable advancements in accuracy, yet the further improvement is restricted by the lagged tracking robustness. We find the underlying reason is that the regression network in anchor-based methods is only trained on the positive anchor boxes (i.e., $IoU \geq0.6$). This mechanism makes it difficult to refine the anchors whose overlap with the target objects are small. In this paper, we propose a novel object-aware anchor-free network to address this issue. First, instead of refining the reference anchor boxes, we directly predict the position and scale of target objects in an anchor-free fashion. Since each pixel in groundtruth boxes is well trained, the tracker is capable of rectifying inexact predictions of target objects during inference. Second, we introduce a feature alignment module to learn an object-aware feature from predicted bounding boxes. The object-aware feature can further contribute to the classification of target objects and background. Moreover, we present a novel tracking framework based on the anchor-free model. The experiments show that our anchor-free tracker achieves state-of-the-art performance on five benchmarks, including VOT-2018, VOT-2019, OTB-100, GOT-10k and LaSOT. The source code is available at https://github.com/researchmm/TracKit.
We study on image super-resolution (SR), which aims to recover realistic textures from a low-resolution (LR) image. Recent progress has been made by taking high-resolution images as references (Ref), so that relevant textures can be transferred to LR images. However, existing SR approaches neglect to use attention mechanisms to transfer high-resolution (HR) textures from Ref images, which limits these approaches in challenging cases. In this paper, we propose a novel Texture Transformer Network for Image Super-Resolution (TTSR), in which the LR and Ref images are formulated as queries and keys in a transformer, respectively. TTSR consists of four closely-related modules optimized for image generation tasks, including a learnable texture extractor by DNN, a relevance embedding module, a hard-attention module for texture transfer, and a soft-attention module for texture synthesis. Such a design encourages joint feature learning across LR and Ref images, in which deep feature correspondences can be discovered by attention, and thus accurate texture features can be transferred. The proposed texture transformer can be further stacked in a cross-scale way, which enables texture recovery from different levels (e.g., from 1x to 4x magnification). Extensive experiments show that TTSR achieves significant improvements over state-of-the-art approaches on both quantitative and qualitative evaluations.
This paper presents a Multitask Multilingual Multimodal Pre-trained model (M3P) that combines multilingual-monomodal pre-training and monolingual-multimodal pre-training into a unified framework via multitask learning and weight sharing. The model learns universal representations that can map objects that occurred in different modalities or expressed in different languages to vectors in a common semantic space. To verify the generalization capability of M3P, we fine-tune the pre-trained model for different types of downstream tasks: multilingual image-text retrieval, multilingual image captioning, multimodal machine translation, multilingual natural language inference and multilingual text generation. Evaluation shows that M3P can (i) achieve comparable results on multilingual tasks and English multimodal tasks, compared to the state-of-the-art models pre-trained for these two types of tasks separately, and (ii) obtain new state-of-the-art results on non-English multimodal tasks in the zero-shot or few-shot setting. We also build a new Multilingual Image-Language Dataset (MILD) by collecting large amounts of (text-query, image, context) triplets in 8 languages from the logs of a commercial search engine
This paper reviews the NTIRE 2020 challenge on perceptual extreme super-resolution with focus on proposed solutions and results. The challenge task was to super-resolve an input image with a magnification factor 16 based on a set of prior examples of low and corresponding high resolution images. The goal is to obtain a network design capable to produce high resolution results with the best perceptual quality and similar to the ground truth. The track had 280 registered participants, and 19 teams submitted the final results. They gauge the state-of-the-art in single image super-resolution.
We propose Pixel-BERT to align image pixels with text by deep multi-modal transformers that jointly learn visual and language embedding in a unified end-to-end framework. We aim to build a more accurate and thorough connection between image pixels and language semantics directly from image and sentence pairs instead of using region-based image features as the most recent vision and language tasks. Our Pixel-BERT which aligns semantic connection in pixel and text level solves the limitation of task-specific visual representation for vision and language tasks. It also relieves the cost of bounding box annotations and overcomes the unbalance between semantic labels in visual task and language semantic. To provide a better representation for down-stream tasks, we pre-train a universal end-to-end model with image and sentence pairs from Visual Genome dataset and MS-COCO dataset. We propose to use a random pixel sampling mechanism to enhance the robustness of visual representation and to apply the Masked Language Model and Image-Text Matching as pre-training tasks. Extensive experiments on downstream tasks with our pre-trained model show that our approach makes the most state-of-the-arts in downstream tasks, including Visual Question Answering (VQA), image-text retrieval, Natural Language for Visual Reasoning for Real (NLVR). Particularly, we boost the performance of a single model in VQA task by 2.17 points compared with SOTA under fair comparison.
In this report, we introduce the Winner method for HACS Temporal Action Localization Challenge 2019. Temporal action localization is challenging since a target proposal may be related to several other candidate proposals in an untrimmed video. Existing methods cannot tackle this challenge well since temporal proposals are considered individually and their temporal dependencies are neglected. To address this issue, we propose sparse 2D temporal adjacent networks to model the temporal relationship between candidate proposals. This method is built upon the recent proposed 2D-TAN approach. The sampling strategy in 2D-TAN introduces the unbalanced context problem, where short proposals can perceive more context than long proposals. Therefore, we further propose a Sparse 2D Temporal Adjacent Network (S-2D-TAN). It is capable of involving more context information for long proposals and further learning discriminative features from them. By combining our S-2D-TAN with a simple action classifier, our method achieves a mAP of 23.49 on the test set, which win the first place in the HACS challenge.