Italy

We present VINNAS: Variational inference-based neural network architecture search for searching sparse convolutional neural networks. Our approach finds the optimal neural architecture by dropping out candidate operations in an overparameterised supergraph using variational dropout with automatic relevance determination priors, which makes the algorithm gradually remove unnecessary operations and connections.

ICPR 2020

VINNAS: Variational inference-based neural network architecture search

convolutional neural networks

bayesian inference

neural architecture search

ICPR 2020 is the premier international conference on Pattern Recognition, gathering leading researchers from around the world to present their latest achievements in Pattern Analysis and Recognition, applied Artificial Intelligence and Machine Learning. The topic of the year is " Putting Artificial intelligence to work on Patterns". 
For the first time, ICPR has employed a two-round review process similar to journal submissions, with new papers allowed to be submitted in either the first or second round and papers submitted in the first round and not accepted allowed to be revised and re-submitted for second round review. 

In this edition of the **ICPR 2020** Conference, there are 1407 Oral/Poster presentations organized in five Tracks: 
- Artificial Intelligence, Machine Learning for Pattern Analysis; 
- Biometrics, Human Analysis and Behavior Understanding; Computer Vision, Robotics and Intelligent Systems; 
- Document and Media Analysis; and 
- Image and Signal Processing. 

The **ICPR 2020** Main Conference program also includes 7 Keynotes from prominent world-renowned scientists, 8 Tutorials, 28 Demos, 8 Challenges, 2 Panels, as well as an Industry/Sponsor Exhibition. In addition, 41 Workshops on a range of important current and emerging research topics complete the ICPR 2020 program. 


poster

For the action localization task, pre-defined action anchors are the cornerstone of mainstream techniques. State-ofthe-art models mostly rely on a dense segmenting scheme, where anchors are sampled uniformly over the temporal domain with a predefined set of scales. However, it is not sufficient because action duration varies greatly. Therefore, it is necessary for the anchors or proposals to have a variable receptive field. In this paper, we propose a method called YOLA (You Ought to Look Around) which includes three parts: 1) a robust backbone SPN-I3D for extracting spatio-temporal features. In this part, we employ a stronger backbone I3D with SPN (Segment Pyramid Network) instead of C3D to obtain multi-scale features; 2) a simple but useful feature fusion module named LFE (Local Feature Extraction). Compared with the fully connected layer and global average pooling, our LFE model is more advantageous for network to fit and fuse features. 3) a new feature segment aligning method called TPGC (Two Pathway Graph Convolution), which allows one proposal to leverage semantic features of adjacent proposals to update its content and make sure the proposals have a variable receptive field. YOLA add only a small overhead to the baseline network, and is easy to train in an end-to-end manner, running at a speed of 1097 fps. YOLA achieves a mAP of 58.3%, outperforming all existing models including both RGBbased and two stream on THUMOS’14, and achieves competitive results on ActivityNet 1.3.

You Ought to Look Around: Precise, Large Span Action Detection

Learning-based stereo matching has recently achieved promising results, yet still suffers difficulties in establishing reliable matches in weakly matchable regions that are textureless, non-Lambertian, or occluded. In this paper, we address this challenge by proposing a stereo matching network that considers pixel-wise matchability. Specifically, the network jointly regresses disparity and matchability maps from 3D probability volume through expectation and entropy operations. Next, a learned attenuation is applied as the robust loss function to alleviate the influence of weakly matchable pixels in the training. Finally, a matchability-aware disparity refinement is introduced to improve the depth inference in weakly matchable regions. The proposed deep stereo matchability (DSM) framework can
improve the matching result or accelerate the computation while still guaranteeing the quality. Moreover, the DSM framework is portable to many recent stereo networks. Extensive experiments are conducted on Scene Flow and KITTI stereo datasets to demonstrate the effectiveness of the proposed framework over the state-of-the-art learning-based stereo methods.

Learning Stereo Matchability in Disparity Regression Networks

The primitive basis of image based material recognition builds upon the fact that discrepancies in the reflectances of distinct materials lead to imaging differences under multiple viewpoints. LF cameras possess coherent abilities to capture multiple sub-aperture views (SAIs) within one exposure, which can provide appropriate multi-view sources for material recognition. In this paper, a unified ``Factorize-Connect-Merge`` (FCM) deep-learning pipeline is proposed to solve problems of light field image based material recognition. 4D light-field data as input is initially decomposed into consecutive 3D light-field slices. Shallow CNN is leveraged to extract low-level visual features of each view inside these slices. As to establish correspondences between these SAIs, Bidirectional Long-Short Term Memory (Bi-LSTM) network is built upon these low-level features to model the imaging differences. After feature selection including concatenation and dimension reduction, effective and robust feature representations for material recognition can be extracted from 4D light-field data. Experimental results indicate that the proposed pipeline can obtain remarkable performances on both tasks of single-pixel material classification and whole-image material segmentation. In addition, the proposed pipeline can potentially benefit and inspire other researchers who may also take LF images as input and need to extract 4D light-field representations for computer vision tasks such as object classification, semantic segmentation and edge detection.

A Novel Deep-learning Pipeline for Light Field Image Based Material Recognition

Object detection in Earth Vision has achieved great progress in recent years. However, tiny object detection in aerial images remains a very challenging problem since the tiny objects contain a small number of pixels and are easily confused with
the background. To advance tiny object detection research in aerial images, we present a new dataset for Tiny Object Detection in Aerial Images (AI-TOD). Specifically, AI-TOD comes with 700,621 object instances for eight categories across 28,036 aerial images. Compared to existing object detection datasets in aerial images, the mean size of objects in AI-TOD is about 12.8 pixels, which is much smaller than others. To build a benchmark for tiny object detection in aerial images, we evaluate the state-of-the-art object detectors on our AI-TOD dataset. Experimental results show that direct application of these approaches on AI-TOD
produces suboptimal object detection results, thus new specialized detectors for tiny object detection need to be designed. Therefore, we propose a multiple center points based learning network (M-CenterNet) to improve the localization performance of tiny object detection, and experimental results show the significant
performance gain over the competitors.

Tiny Object Detection in Aerial Images

Current research on embodied AI mainly focuses on stand-alone tasks. Instead, we aim at bridging recent findings on embodied exploration and image captioning.
We devise a novel setting involving the exploration of a photorealistic indoor environment while describing the current RGB observation, and we call it Explore and Explain.
We present a novel solution involving curiosity-driven exploration and we apply a captioning algorithm to describe relevant objects along the trajectory of the agent.

Explore and Explain: Self-Supervised Navigation and Recounting

Mesh refinement is a fundamental step for accurate Multi-View Stereo. It modifies the geometry of an initial manifold mesh to minimize the photometric error induced in a set of camera pairs. This initial mesh is usually the output of volumetric 3D reconstruction based on min-cut over Delaunay Triangulations. Such methods produce a significant amount of non-manifold vertices, therefore they require a vertex split step to explicitly repair them. In this paper, we extend this method to preemptively fix the non-manifold vertices by reasoning directly on the Delaunay Triangulation and avoid most vertex splits. The main contribution of this paper addresses the problem of choosing the camera pairs adopted by the refinement process. We treat the problem as a mesh labeling process, where each label corresponds to a camera pair. Differently from the state-of-the-art methods, which use each camera pair to refine all the visible parts of the mesh, we choose, for each facet, the best pair that enforces both the overall visibility and coverage. The refinement step is applied for each facet using only the camera pair selected. This facetwise refinement helps the process to be applied in the most evenly way possible.

Facetwise Mesh Refinement for Multi-View Stereo

The application of deep learning to 3D point clouds is challenging due to its lack of order. Inspired by the point embeddings of PointNet and the edge embeddings of DGCNNs, we propose three improvements to the task of point cloud analysis. First, we introduce a novel feature-attentive neural network layer, a FAT layer, that combines both global point-based features and local edge-based features in order to generate better embeddings. Second, we find that applying the same attention mechanism across two different forms of feature map aggregation, max pooling and average pooling, gives better performance than either alone. Third, we observe that residual feature reuse in this setting propagates information more effectively between the layers, and makes the network easier to train. Our architecture achieves state-of-the-art results on the task of point cloud classification, as demonstrated on the ModelNet40 dataset, and an extremely competitive performance on the ShapeNet part segmentation challenge.

FatNet: A Feature-attentive Network for 3D Point Cloud Processing

Activity detection in security videos is a difficult problem due to multiple factors such as large field of view, presence of multiple activities, varying scales and viewpoints, and its untrimmed nature. The existing research in activity detection is mainly focused on datasets, such as UCF-101, JHMDB, THUMOS, and AVA, which partially address these issues. The requirement of processing the security videos in realtime makes this even more challenging. In this work we propose Gabriella, a real-time online system to perform activity detection on untrimmed security videos. The proposed method consists of three stages: tubelet extraction, activity classification, and online tubelet merging. For tubelet extraction, we propose a localization network which takes a video clip as input and spatio-temporally detects potential foreground regions at multiple scales to generate action tubelets. We propose a novel Patch-Dice loss to handle large variations in actor size. Our online processing of videos at a clip level drastically reduces the computation time in detecting activities. The detected tubelets are assigned activity class scores by the classification network and merged together using our proposed Tubelet-Merge Action-Split (TMAS) algorithm to form the final action detections. The TMAS algorithm efficiently connects the tubelets in an online fashion to generate action detections which are robust against varying length activities. We perform our experiments on the VIRAT and MEVA (Multiview Extended Video with Activities) datasets and demonstrate the effectiveness of the proposed approach in terms of speed (∼100 fps) and performance with state-of-the-art results.

Gabriella: An Online System for Real-Time Activity Detection in Untrimmed Security Videos

Mass utilization of body-worn cameras has led to a huge corpus of available egocentric video. Existing video summarization algorithms can accelerate browsing such videos by selecting (visually) interesting shots from them. Nonetheless, since the system user still has to watch the summary videos, browsing large video databases remain a challenge. Hence, in this work, we propose to generate a textual synopsis, consisting of a few sentences describing the most important events in a long egocentric videos. Users can read the short text to gain insight about the video, and more importantly, efficiently search through the content of a large video database using text queries. Since egocentric videos are long and contain many activities and events, using video-to-text algorithms results in thousands of descriptions, many of which are incorrect. Therefore, we propose a multi-task learning scheme to simultaneously generate descriptions for video segments and summarize the resulting descriptions in an end-to-end fashion. We Input a set of video shots and the network generates a text description for each shot. Next, visual-language content matching unit that is trained with a weakly supervised objective, identifies the correct descriptions. Finally, the last component of our network, called purport network, evaluates the descriptions all together to select the ones containing crucial information. Out of thousands of descriptions generated for the video, a few informative sentences are returned to the user. We validate our framework on the challenging UT Egocentric video dataset, where each video is between 3 to 5 hours long, associated with over 3000 textual descriptions on average. The generated textual summaries, including only 5 percent (or less) of the generated descriptions, are compared to groundtruth summaries in text domain using well-established metrics in natural language processing.

Text Synopsis Generation for Egocentric Videos

Nowadays 360 video analysis has become a significant research topic in the field since the appearance of high-quality and low-cost 360 wearable devices. In this paper, we propose a novel LiteFlowNet360 architecture for 360 videos optical flow estimation. We design LiteFlowNet360 as a domain adaptation framework from perspective video domain to 360 video domain. We adapt it from simple kernel transformation techniques inspired by Kernel Transformer Network (KTN) to cope with inherent distortion in 360 videos caused by the sphere-to-plane projection. First, we apply an incremental transformation of convolution layers in feature pyramid network and show that further transformation in inference and regularization layers are not important, hence reducing the network growth in terms of size and computation cost. Second, we refine the network by training with augmented data in a supervised manner. We perform data augmentation by projecting the images in a sphere and re-projecting to a plane. Third, we train LiteFlowNet360 in a self-supervised manner using target domain 360 videos. Experimental results show the promising results of 360 video optical flow estimation using the proposed novel architecture.

Premium content

Downloads

Next from ICPR 2020

You Ought to Look Around: Precise, Large Span Action Detection

Similar lecture

VINNAS: Variational inference-based neural network architecture search