Zhen He

Xiaochuan Fan, Chi Zhang, Yong Yang et al.

Product images are essential for providing desirable user experience in an e-commerce platform. For a platform with billions of products, it is extremely time-costly and labor-expensive to manually pick and organize qualified images. Furthermore, there are the numerous and complicated image rules that a product image needs to comply in order to be generated/selected. To address these challenges, in this paper, we present a new learning framework in order to achieve Automatic Generation of Product-Image Sequence (AGPIS) in e-commerce. To this end, we propose a Multi-modality Unified Image-sequence Classifier (MUIsC), which is able to simultaneously detect all categories of rule violations through learning. MUIsC leverages textual review feedback as the additional training target and utilizes product textual description to provide extra semantic information. Based on offline evaluations, we show that the proposed MUIsC significantly outperforms various baselines. Besides MUIsC, we also integrate some other important modules in the proposed framework, such as primary image selection, noncompliant content detection, and image deduplication. With all these modules, our framework works effectively and efficiently in JD.com recommendation platform. By Dec 2021, our AGPIS framework has generated high-standard images for about 1.5 million products and achieves 13.6% in reject rate.

Lin Yuan, Zhen He, Qiang Wang et al.

Human action recognition is a quite hugely investigated area where most remarkable action recognition networks usually use large-scale coarse-grained action datasets of daily human actions as inputs to state the superiority of their networks. We intend to recognize our small-scale fine-grained Tai Chi action dataset using neural networks and propose a transfer-learning method using NTU RGB+D dataset to pre-train our network. More specifically, the proposed method first uses a large-scale NTU RGB+D dataset to pre-train the Transformer-based network for action recognition to extract common features among human motion. Then we freeze the network weights except for the fully connected (FC) layer and take our Tai Chi actions as inputs only to train the initialized FC weights. Experimental results show that our general model pipeline can reach a high accuracy of small-scale fine-grained Tai Chi action recognition with even few inputs and demonstrate that our method achieves the state-of-the-art performance compared with previous Tai Chi action recognition methods.

Xueying Zhang, Kai Shen, Chi Zhang et al.

In this paper, we proposed an automatic Scenario-based Multi-product Advertising Copywriting Generation system (SMPACG) for E-Commerce, which has been deployed on a leading Chinese e-commerce platform. The proposed SMPACG consists of two main components: 1) an automatic multi-product combination selection module, which itself is consisted of a topic prediction model, a pattern and attribute-based selection model and an arbitrator model; and 2) an automatic multi-product advertising copywriting generation module, which combines our proposed domain-specific pretrained language model and knowledge-based data enhancement model. The SMPACG is the first system that realizes automatic scenario-based multi-product advertising contents generation, which achieves significant improvements over other state-of-the-art methods. The SMPACG has been not only developed for directly serving for our e-commerce recommendation system, but also used as a real-time writing assistant tool for merchants.

Jiayang Bai, Zhen He, Shan Yang et al.

Predicting panoramic indoor lighting from a single perspective image is a fundamental but highly ill-posed problem in computer vision and graphics. To achieve locale-aware and robust prediction, this problem can be decomposed into three sub-tasks: depth-based image warping, panorama inpainting and high-dynamic-range (HDR) reconstruction, among which the success of panorama inpainting plays a key role. Recent methods mostly rely on convolutional neural networks (CNNs) to fill the missing contents in the warped panorama. However, they usually achieve suboptimal performance since the missing contents occupy a very large portion in the panoramic space while CNNs are plagued by limited receptive fields. The spatially-varying distortion in the spherical signals further increases the difficulty for conventional CNNs. To address these issues, we propose a local-to-global strategy for large-scale panorama inpainting. In our method, a depth-guided local inpainting is first applied on the warped panorama to fill small but dense holes. Then, a transformer-based network, dubbed PanoTransformer, is designed to hallucinate reasonable global structures in the large holes. To avoid distortion, we further employ cubemap projection in our design of PanoTransformer. The high-quality panorama recovered at any locale helps us to capture spatially-varying indoor illumination with physically-plausible global structures and fine details.

Brandon Victor, Mathilde Letard, Peter Naylor et al.

Floods are among the most common and devastating natural hazards, imposing immense costs on our society and economy due to their disastrous consequences. Recent progress in weather prediction and spaceborne flood mapping demonstrated the feasibility of anticipating extreme events and reliably detecting their catastrophic effects afterwards. However, these efforts are rarely linked to one another and there is a critical lack of datasets and benchmarks to enable the direct forecasting of flood extent. To resolve this issue, we curate a novel dataset enabling a timely prediction of flood extent. Furthermore, we provide a representative evaluation of state-of-the-art methods, structured into two benchmark tracks for forecasting flood inundation maps i) in general and ii) focused on coastal regions. Altogether, our dataset and benchmark provide a comprehensive platform for evaluating flood forecasts, enabling future solutions for this critical challenge. Data, code & models are shared at https://github.com/Multihuntr/GFF under a CC0 license.

Brandon Victor, Zhen He, Aiden Nibali

Agricultural research is essential for increasing food production to meet the requirements of an increasing population in the coming decades. Recently, satellite technology has been improving rapidly and deep learning has seen much success in generic computer vision tasks and many application areas which presents an important opportunity to improve analysis of agricultural land. Here we present a systematic review of 150 studies to find the current uses of deep learning on satellite imagery for agricultural research. Although we identify 5 categories of agricultural monitoring tasks, the majority of the research interest is in crop segmentation and yield prediction. We found that, when used, modern deep learning methods consistently outperformed traditional machine learning across most tasks; the only exception was that Long Short-Term Memory (LSTM) Recurrent Neural Networks did not consistently outperform Random Forests (RF) for yield prediction. The reviewed studies have largely adopted methodologies from generic computer vision, except for one major omission: benchmark datasets are not utilised to evaluate models across studies, making it difficult to compare results. Additionally, some studies have specifically utilised the extra spectral resolution available in satellite imagery, but other divergent properties of satellite images - such as the hugely different scales of spatial patterns - are not being taken advantage of in the reviewed studies.

Haitao Wu, Qing Li, Changqing Zhang et al.

Can our brain signals faithfully reflect the original visual stimuli, even including high-frequency details? Although human perceptual and cognitive capacities enable us to process and remember visual information, these abilities are constrained by several factors, such as limited attentional resources and the finite capacity of visual memory. When visual stimuli are processed by human visual system into brain signals, some information is inevitably lost, leading to a discrepancy known as the \textbf{System GAP}. Additionally, perceptual and cognitive dynamics, along with technical noise in signal acquisition, degrade the fidelity of brain signals relative to the visual stimuli, known as the \textbf{Random GAP}. When encoded brain representations are directly aligned with the corresponding pretrained image features, the System GAP and Random GAP between paired data challenge the model, requiring it to bridge these gaps. However, in the context of limited paired data, these gaps are difficult for the model to learn, leading to overfitting and poor generalization to new data. To address these GAPs, we propose a simple yet effective approach called the \textbf{Uncertainty-aware Blur Prior (UBP)}. It estimates the uncertainty within the paired data, reflecting the mismatch between brain signals and visual stimuli. Based on this uncertainty, UBP dynamically blurs the high-frequency details of the original images, reducing the impact of the mismatch and improving alignment. Our method achieves a top-1 accuracy of \textbf{50.9\%} and a top-5 accuracy of \textbf{79.7\%} on the zero-shot brain-to-image retrieval task, surpassing previous state-of-the-art methods by margins of \textbf{13.7\%} and \textbf{9.8\%}, respectively. Code is available at \href{https://github.com/HaitaoWuTJU/Uncertainty-aware-Blur-Prior}{GitHub}.

Long Nguyen, Aiden Nibali, Joshua Millward et al.

Recently there have been many algorithms proposed for the classification of very high resolution whole slide images (WSIs). These new algorithms are mostly focused on finding novel ways to combine the information from small local patches extracted from the slide, with an emphasis on effectively aggregating more global information for the final predictor. In this paper we thoroughly explore different key design choices for WSI classification algorithms to investigate what matters most for achieving high accuracy. Surprisingly, we found that capturing global context information does not necessarily mean better performance. A model that captures the most global information consistently performs worse than a model that captures less global information. In addition, a very simple multi-instance learning method that captures no global information performs almost as well as models that capture a lot of global information. These results suggest that the most important features for effective WSI classification are captured at the local small patch level, where cell and tissue micro-environment detail is most pronounced. Another surprising finding was that unsupervised pre-training on a larger set of 33 cancers gives significantly worse performance compared to pre-training on a smaller dataset of 7 cancers (including the target cancer). We posit that pre-training on a smaller, more focused dataset allows the feature extractor to make better use of the limited feature space to better discriminate between subtle differences in the input patch.

Zhen He, Jian Li, Daxue Liu et al.

Online Multi-Object Tracking (MOT) from videos is a challenging computer vision task which has been extensively studied for decades. Most of the existing MOT algorithms are based on the Tracking-by-Detection (TBD) paradigm combined with popular machine learning approaches which largely reduce the human effort to tune algorithm parameters. However, the commonly used supervised learning approaches require the labeled data (e.g., bounding boxes), which is expensive for videos. Also, the TBD framework is usually suboptimal since it is not end-to-end, i.e., it considers the task as detection and tracking, but not jointly. To achieve both label-free and end-to-end learning of MOT, we propose a Tracking-by-Animation framework, where a differentiable neural model first tracks objects from input frames and then animates these objects into reconstructed frames. Learning is then driven by the reconstruction error through backpropagation. We further propose a Reprioritized Attentive Tracking to improve the robustness of data association. Experiments conducted on both synthetic and real video datasets show the potential of the proposed model. Our project page is publicly available at: https://github.com/zhen-he/tracking-by-animation

Haritha Thilakarathne, Aiden Nibali, Zhen He et al.

Group activity recognition in video is a complex task due to the need for a model to recognise the actions of all individuals in the video and their complex interactions. Recent studies propose that optimal performance is achieved by individually tracking each person and subsequently inputting the sequence of poses or cropped images/optical flow into a model. This helps the model to recognise what actions each person is performing before they are merged to arrive at the group action class. However, all previous models are highly reliant on high quality tracking and have only been evaluated using ground truth tracking information. In practice it is almost impossible to achieve highly reliable tracking information for all individuals in a group activity video. We introduce an innovative deep learning-based group activity recognition approach called Rendered Pose based Group Activity Recognition System (RePGARS) which is designed to be tolerant of unreliable tracking and pose information. Experimental results confirm that RePGARS outperforms all existing group activity recognition algorithms tested which do not use ground truth detection and tracking information.

JaeWoong Shin, Jeongun Ryu, Aaron Valero Puche et al.

Pathologists routinely alternate between different magnifications when examining Whole-Slide Images, allowing them to evaluate both broad tissue morphology and intricate cellular details to form comprehensive diagnoses. However, existing deep learning-based cell detection models struggle to replicate these behaviors and learn the interdependent semantics between structures at different magnifications. A key barrier in the field is the lack of datasets with multi-scale overlapping cell and tissue annotations. The OCELOT 2023 challenge was initiated to gather insights from the community to validate the hypothesis that understanding cell and tissue (cell-tissue) interactions is crucial for achieving human-level performance, and to accelerate the research in this field. The challenge dataset includes overlapping cell detection and tissue segmentation annotations from six organs, comprising 673 pairs sourced from 306 The Cancer Genome Atlas (TCGA) Whole-Slide Images with hematoxylin and eosin staining, divided into training, validation, and test subsets. Participants presented models that significantly enhanced the understanding of cell-tissue relationships. Top entries achieved up to a 7.99 increase in F1-score on the test set compared to the baseline cell-only model that did not incorporate cell-tissue relationships. This is a substantial improvement in performance over traditional cell-only detection methods, demonstrating the need for incorporating multi-scale semantics into the models. This paper provides a comparative analysis of the methods used by participants, highlighting innovative strategies implemented in the OCELOT 2023 challenge.

Jiayang Bai, Jie Guo, Chenchen Wan et al.

Lighting prediction from a single image is becoming increasingly important in many vision and augmented reality (AR) applications in which shading and shadow consistency between virtual and real objects should be guaranteed. However, this is a notoriously ill-posed problem, especially for indoor scenarios, because of the complexity of indoor luminaires and the limited information involved in 2D images. In this paper, we propose a graph learning-based framework for indoor lighting estimation. At its core is a new lighting model (dubbed DSGLight) based on depth-augmented Spherical Gaussians (SG) and a Graph Convolutional Network (GCN) that infers the new lighting representation from a single LDR image of limited field-of-view. Our lighting model builds 128 evenly distributed SGs over the indoor panorama, where each SG encoding the lighting and the depth around that node. The proposed GCN then learns the mapping from the input image to DSGLight. Compared with existing lighting models, our DSGLight encodes both direct lighting and indirect environmental lighting more faithfully and compactly. It also makes network training and inference more stable. The estimated depth distribution enables temporally stable shading and shadows under spatially-varying lighting. Through thorough experiments, we show that our method obviously outperforms existing methods both qualitatively and quantitatively.

Xiaojie Guo, Shugen Wang, Hanqing Zhao et al.

In the past decade, automatic product description generation for e-commerce have witnessed significant advancement. As the services provided by e-commerce platforms become diverse, it is necessary to dynamically adapt the patterns of descriptions generated. The selling point of products is an important type of product description for which the length should be as short as possible while still conveying key information. In addition, this kind of product description should be eye-catching to the readers. Currently, product selling points are normally written by human experts. Thus, the creation and maintenance of these contents incur high costs. These costs can be significantly reduced if product selling points can be automatically generated by machines. In this paper, we report our experience developing and deploying the Intelligent Online Selling Point Extraction (IOSPE) system to serve the recommendation system in the JD.com e-commerce platform. Since July 2020, IOSPE has become a core service for 62 key categories of products (covering more than 4 million products). So far, it has generated more than 0.1 billion selling points, thereby significantly scaling up the selling point creation operation and saving human labour. These IOSPE generated selling points have increased the click-through rate (CTR) by 1.89\% and the average duration the customers spent on the products by more than 2.03\% compared to the previous practice, which are significant improvements for such a large-scale e-commerce platform.

Xueying Zhang, Yanyan Zou, Hainan Zhang et al.

Product copywriting is a critical component of e-commerce recommendation platforms. It aims to attract users' interest and improve user experience by highlighting product characteristics with textual descriptions. In this paper, we report our experience deploying the proposed Automatic Product Copywriting Generation (APCG) system into the JD.com e-commerce product recommendation platform. It consists of two main components: 1) natural language generation, which is built from a transformer-pointer network and a pre-trained sequence-to-sequence model based on millions of training data from our in-house platform; and 2) copywriting quality control, which is based on both automatic evaluation and human screening. For selected domains, the models are trained and updated daily with the updated training data. In addition, the model is also used as a real-time writing assistant tool on our live broadcast platform. The APCG system has been deployed in JD.com since Feb 2021. By Sep 2021, it has generated 2.53 million product descriptions, and improved the overall averaged click-through rate (CTR) and the Conversion Rate (CVR) by 4.22% and 3.61%, compared to baselines, respectively on a year-on-year basis. The accumulated Gross Merchandise Volume (GMV) made by our system is improved by 213.42%, compared to the number in Feb 2021.

Tri Huynh, Aiden Nibali, Zhen He

Medical image classification is often challenging for two reasons: a lack of labelled examples due to expensive and time-consuming annotation protocols, and imbalanced class labels due to the relative scarcity of disease-positive individuals in the wider population. Semi-supervised learning (SSL) methods exist for dealing with a lack of labels, but they generally do not address the problem of class imbalance. In this study we propose Adaptive Blended Consistency Loss (ABCL), a drop-in replacement for consistency loss in perturbation-based SSL methods. ABCL counteracts data skew by adaptively mixing the target class distribution of the consistency loss in accordance with class frequency. Our experiments with ABCL reveal improvements to unweighted average recall on two different imbalanced medical image classification datasets when compared with existing consistency losses that are not designed to counteract class imbalance.

Haritha Thilakarathne, Aiden Nibali, Zhen He et al.

We introduce a novel deep learning based group activity recognition approach called the Pose Only Group Activity Recognition System (POGARS), designed to use only tracked poses of people to predict the performed group activity. In contrast to existing approaches for group activity recognition, POGARS uses 1D CNNs to learn spatiotemporal dynamics of individuals involved in a group activity and forgo learning features from pixel data. The proposed model uses a spatial and temporal attention mechanism to infer person-wise importance and multi-task learning for simultaneously performing group and individual action classification. Experimental results confirm that POGARS achieves highly competitive results compared to state-of-the-art methods on a widely used public volleyball dataset despite only using tracked pose as input. Further our experiments show by using pose only as input, POGARS has better generalization capabilities compared to methods that use RGB as input.

Xu Yuan, Hongshen Chen, Yonghao Song et al.

Most sequential recommendation models capture the features of consecutive items in a user-item interaction history. Though effective, their representation expressiveness is still hindered by the sparse learning signals. As a result, the sequential recommender is prone to make inconsistent predictions. In this paper, we propose a model, SSI, to improve sequential recommendation consistency with Self-Supervised Imitation. Precisely, we extract the consistency knowledge by utilizing three self-supervised pre-training tasks, where temporal consistency and persona consistency capture user-interaction dynamics in terms of the chronological order and persona sensitivities, respectively. Furthermore, to provide the model with a global perspective, global session consistency is introduced by maximizing the mutual information among global and local interaction sequences. Finally, to comprehensively take advantage of all three independent aspects of consistency-enhanced knowledge, we establish an integrated imitation learning framework. The consistency knowledge is effectively internalized and transferred to the student model by imitating the conventional prediction logit as well as the consistency-enhanced item representations. In addition, the flexible self-supervised imitation framework can also benefit other student recommenders. Experiments on four real-world datasets show that SSI effectively outperforms the state-of-the-art sequential recommendation methods.

Kristen Moore, Shenjun Zhong, Zhen He et al.

In this paper we present the results of our experiments in training and deploying a self-supervised retrieval-based chatbot trained with contrastive learning for assisting customer support agents. In contrast to most existing research papers in this area where the focus is on solving just one component of a deployable chatbot, we present an end-to-end set of solutions to take the reader from an unlabelled chatlogs to a deployed chatbot. This set of solutions includes creating a self-supervised dataset and a weakly labelled dataset from chatlogs, as well as a systematic approach to selecting a fixed list of canned responses. We present a hierarchical-based RNN architecture for the response selection model, chosen for its ability to cache intermediate utterance embeddings, which helped to meet deployment inference speed requirements. We compare the performance of this architecture across 3 different learning objectives: self-supervised contrastive learning, binary classification, and multi-class classification. We find that using a self-supervised contrastive learning model outperforms training the binary and multi-class classification models on a weakly labelled dataset. Our results validate that the self-supervised contrastive learning approach can be effectively used for a real-world chatbot scenario.

Brandon Victor, Aiden Nibali, Zhen He et al.

Sophisticated trajectory prediction models that effectively mimic team dynamics have many potential uses for sports coaches, broadcasters and spectators. However, through experiments on soccer data we found that it can be surprisingly challenging to train a deep learning model for player trajectory prediction which outperforms linear extrapolation on average distance between predicted and true future trajectories. We propose and test a novel method for improving training by predicting a sparse trajectory and interpolating using constant acceleration, which improves performance for several models. This interpolation can also be used on models that aren't trained with sparse outputs, and we find that this consistently improves performance for all tested models. Additionally, we find that the accuracy of predicted trajectories for a subset of players can be improved by conditioning on the full trajectories of the other players, and that this is further improved when combined with sparse predictions. We also propose a novel architecture using graph networks and multi-head attention (GraN-MA) which achieves better performance than other tested state-of-the-art models on our dataset and is trivially adapted for both sparse trajectories and full-trajectory conditioned trajectory prediction.

Matthias Langer, Zhen He, Wenny Rahayu et al.

Distributed deep learning systems (DDLS) train deep neural network models by utilizing the distributed resources of a cluster. Developers of DDLS are required to make many decisions to process their particular workloads in their chosen environment efficiently. The advent of GPU-based deep learning, the ever-increasing size of datasets and deep neural network models, in combination with the bandwidth constraints that exist in cluster environments require developers of DDLS to be innovative in order to train high quality models quickly. Comparing DDLS side-by-side is difficult due to their extensive feature lists and architectural deviations. We aim to shine some light on the fundamental principles that are at work when training deep neural networks in a cluster of independent machines by analyzing the general properties associated with training deep learning models and how such workloads can be distributed in a cluster to achieve collaborative model training. Thereby we provide an overview of the different techniques that are used by contemporary DDLS and discuss their influence and implications on the training process. To conceptualize and compare DDLS, we group different techniques into categories, thus establishing a taxonomy of distributed deep learning systems.

Aiden Nibali, Zhen He, Stuart Morgan et al.

Automatically determining three-dimensional human pose from monocular RGB image data is a challenging problem. The two-dimensional nature of the input results in intrinsic ambiguities which make inferring depth particularly difficult. Recently, researchers have demonstrated that the flexible statistical modelling capabilities of deep neural networks are sufficient to make such inferences with reasonable accuracy. However, many of these models use coordinate output techniques which are memory-intensive, not differentiable, and/or do not spatially generalise well. We propose improvements to 3D coordinate prediction which avoid the aforementioned undesirable traits by predicting 2D marginal heatmaps under an augmented soft-argmax scheme. Our resulting model, MargiPose, produces visually coherent heatmaps whilst maintaining differentiability. We are also able to achieve state-of-the-art accuracy on publicly available 3D human pose estimation data.

Aiden Nibali, Zhen He, Stuart Morgan et al.

We study deep learning approaches to inferring numerical coordinates for points of interest in an input image. Existing convolutional neural network-based solutions to this problem either take a heatmap matching approach or regress to coordinates with a fully connected output layer. Neither of these approaches is ideal, since the former is not entirely differentiable, and the latter lacks inherent spatial generalization. We propose our differentiable spatial to numerical transform (DSNT) to fill this gap. The DSNT layer adds no trainable parameters, is fully differentiable, and exhibits good spatial generalization. Unlike heatmap matching, DSNT works well with low heatmap resolutions, so it can be dropped in as an output layer for a wide range of existing fully convolutional architectures. Consequently, DSNT offers a better trade-off between inference speed and prediction accuracy compared to existing techniques. When used to replace the popular heatmap matching approach used in almost all state-of-the-art methods for pose estimation, DSNT gives better prediction accuracy for all model architectures tested.

Zhen He, Shaobing Gao, Liang Xiao et al.

Long Short-Term Memory (LSTM) is a popular approach to boosting the ability of Recurrent Neural Networks to store longer term temporal information. The capacity of an LSTM network can be increased by widening and adding layers. However, usually the former introduces additional parameters, while the latter increases the runtime. As an alternative we propose the Tensorized LSTM in which the hidden states are represented by tensors and updated via a cross-layer convolution. By increasing the tensor size, the network can be widened efficiently without additional parameters since the parameters are shared across different locations in the tensor; by delaying the output, the network can be deepened implicitly with little additional runtime since deep computations for each timestep are merged into temporal computations of the sequence. Experiments conducted on five challenging sequence learning tasks show the potential of the proposed model.

Brandon Victor, Zhen He, Stuart Morgan et al.

In many sports, it is useful to analyse video of an athlete in competition for training purposes. In swimming, stroke rate is a common metric used by coaches; requiring a laborious labelling of each individual stroke. We show that using a Convolutional Neural Network (CNN) we can automatically detect discrete events in continuous video (in this case, swimming strokes). We create a CNN that learns a mapping from a window of frames to a point on a smooth 1D target signal, with peaks denoting the location of a stroke, evaluated as a sliding window. To our knowledge this process of training and utilizing a CNN has not been investigated before; either in sports or fundamental computer vision research. Most research has been focused on action recognition and using it to classify many clips in continuous video for action localisation. In this paper we demonstrate our process works well on the task of detecting swimming strokes in the wild. However, without modifying the model architecture or training method, the process is also shown to work equally well on detecting tennis strokes, implying that this is a general process. The outputs of our system are surprisingly smooth signals that predict an arbitrary event at least as accurately as humans (manually evaluated from a sample of negative results). A number of different architectures are evaluated, pertaining to slightly different problem formulations and signal targets.

Aiden Nibali, Zhen He, Stuart Morgan et al.

Due to recent advances in technology, the recording and analysis of video data has become an increasingly common component of athlete training programmes. Today it is incredibly easy and affordable to set up a fixed camera and record athletes in a wide range of sports, such as diving, gymnastics, golf, tennis, etc. However, the manual analysis of the obtained footage is a time-consuming task which involves isolating actions of interest and categorizing them using domain-specific knowledge. In order to automate this kind of task, three challenging sub-problems are often encountered: 1) temporally cropping events/actions of interest from continuous video; 2) tracking the object of interest; and 3) classifying the events/actions of interest. Most previous work has focused on solving just one of the above sub-problems in isolation. In contrast, this paper provides a complete solution to the overall action monitoring task in the context of a challenging real-world exemplar. Specifically, we address the problem of diving classification. This is a challenging problem since the person (diver) of interest typically occupies fewer than 1% of the pixels in each frame. The model is required to learn the temporal boundaries of a dive, even though other divers and bystanders may be in view. Finally, the model must be sensitive to subtle changes in body pose over a large number of frames to determine the classification code. We provide effective solutions to each of the sub-problems which combine to provide a highly functional solution to the task as a whole. The techniques proposed can be easily generalized to video footage recorded from other sports.