Cette thèse, réalisée en coopération avec l’ONERA, concerne la reconnaissance active d’objets 3D par un agent autonome muni d’une caméra d’observation. Alors qu’en reconnaissance passive les modalités d’acquisitions des observations sont imposées et génèrent parfois des ambiguïtés, la reconnaissance active exploite la possibilité de contrôler en ligne ces modalités d’acquisition au cours d’un processus d’inférence séquentiel dans le but de lever l’ambiguïté. L’objectif des travaux est d’établir des stratégies de planification dans l’acquisition de l’information avec le souci d’une mise en œuvre réaliste de la reconnaissance active. Le cadre de l’apprentissage statistique est pour cela mis à profit. La première partie des travaux se consacre à apprendre à planifier. Deux contraintes réalistes sont prise en compte : d’une part, une modélisation imparfaite des objets susceptible de générer des ambiguïtés supplémentaires - d’autre part, le budget d’apprentissage est coûteux (en temps, en énergie), donc limité. La deuxième partie des travaux s’attache à exploiter au mieux les observations au cours de la reconnaissance. La possibilité d’une reconnaissance active multi-échelles est étudiée pour permettre une interprétation au plus tôt dans le processus séquentiel d’acquisition de l’information. Les observations sont également utilisées pour estimer la pose de l’objet de manière robuste afin d’assurer la cohérence entre les modalités planifiées et celles réellement atteintes par l’agent visuel. / This PhD thesis, conducted in cooperation with ONERA, focuses on active 3D object recognition by an autonomous visual agent. Whereas in passive recognition, acquisition modalities of observations are fixed and may generate ambiguities, active recognition exploits the possibility of controling these modalities online in a sequential inference process in order to remove these ambiguities. The aim of this work is to design, in a statistical learning framework, planning strategies in the acquisition of information while achieving a realistic implementation of active recognition. The first part of the work is dedicated to learning to plan. Two realistic constraints are taken into account : on the one hand, planning with imperfect object modeling may generate further ambiguities - on the other hand, the learning cost (in time, energy) is expensive and therefore limited. The second part of this work focuses on maximally exploiting observations acquired during recognition. The possibility of an active multi-scale recognition is investigated to allow an interpretation as soon as the sequential acquisition process begins. Observations are also used to robustly estimate the pose of the object to ensure consistency between the planned and actual modality of the visual agent.

In this thesis we study a perception problem in the context of autonomous driving. Specifically, we study the computer vision problem of 3D object detection, in which objects should be detected from various sensor data and their position in the 3D world should be estimated. We also study the application of Generative Adversarial Networks in domain adaptation techniques, aiming to improve the 3D object detection model's ability to transfer between different domains. The state-of-the-art Frustum-PointNet architecture for LiDAR-based 3D object detection was implemented and found to closely match its reported performance when trained and evaluated on the KITTI dataset. The architecture was also found to transfer reasonably well from the synthetic SYN dataset to KITTI, and is thus believed to be usable in a semi-automatic 3D bounding box annotation process. The Frustum-PointNet architecture was also extended to explicitly utilize image features, which surprisingly degraded its detection performance. Furthermore, an image-only 3D object detection model was designed and implemented, which was found to compare quite favourably with current state-of-the-art in terms of detection performance. Additionally, the PixelDA approach was adopted and successfully applied to the MNIST to MNIST-M domain adaptation problem, which validated the idea that unsupervised domain adaptation using Generative Adversarial Networks can improve the performance of a task network for a dataset lacking ground truth annotations. Surprisingly, the approach did however not significantly improve upon the performance of the image-based 3D object detection models when trained on the SYN dataset and evaluated on KITTI.

Machine Translation Using Syntactic Analysis Martin Popel This thesis describes our improvement of machine translation (MT), with a special focus on the English-Czech language pair, but using techniques ap- plicable also to other languages. First, we present multiple improvements of the deep-syntactic system TectoMT. For instance, we implemented a novel context-sensitive translation model, comparing several machine learning ap- proaches. We also adapted TectoMT to other domains and languages. Sec- ond, we present Transformer - a state-of-the-art end-to-end neural MT sys- tem. We analyzed in detail the effect of several training hyper-parameters. With our optimized training, the system outperformed the best result on the WMT2017 test set by +1.0 BLEU. We further extended this system by uti- lization of monolingual training data and by a new type of backtranslation (+2.8 BLEU compared to the baseline system). In addition, we leveraged domain adaptation and the effect of "translationese" (i.e which language in parallel data is the original and which is the translation) to optimize MT systems for original-language and translated-language data (gaining further +0.2 BLEU). Our improved neural MT system significantly (p¡0.05) out- performed all other systems in English-Czech and Czech-English WMT2018 shared tasks,...

Deep learning works best with vast andd well-distributed data collections. However, collecting and annotating large data sets can be very time-consuming and expensive. Moreover, deep learning is specific to domain knowledge, even with data and computation. E.g., models trained to classify animals would probably underperform when they classify vehicles. Although techniques such as domain adaptation and transfer learning have been popularised recently, tasks in cross-domain knowledge transfer have also taken off. However, most of these works are limited to computer vision. In the domain of time series, this is relatively underexplored. This thesis explores methods to use time series data from one domain to classify data generated from another domain via transfer learning. It focuses on using accelerometer data from running recordings to improve the classification performance on jumping data based on the apparent similarity of individual recordings. Thus, transfer learning and domain adaptation techniques were used to use the learning acquired through deep model training on running sequences. This thesis has performed four experiments to test this domain similarity. The first one consists of transforming time series with the continuous wavelet transform to get both time and frequency information. The model is then pre-trained within a contrastive learning framework. However, the continuous wavelet transformation (CWT) did not improve the classification results. The following two experiments consisted of pre-training the models with self-supervised learning. The first one with a contrastive pretext-task improved the classification results, and the resilience to data decrease. The second one with a forward forecasting pretext-task improved the results when all the data was available but was very sensitive to data decrease. Finally, the domain adaptation was tested and showed interesting performances on the classification task. Although some of the employed techniques did not show improvement, pre-training using contrastive learning on the running dataset has shown great improvement to classify the jumping dataset. / Djupinlärning fungerar bäst med stora och väl distribuerade datasamlingar. Det kan dock vara mycket tidskrävande och dyrt att samla in och kommentera stora datamängder. Även med alla data och beräkningar är djupinlärning specifik för domänkunskap. Exempelvis skulle modeller som tränats för att klassificera djur förmodligen underprestera när de klassificerar fordon. Även om tekniker som domänanpassning och överföringsinlärning har populariserats på senare tid, har även uppgifter inom kunskapsöverföring mellan olika domäner tagit fart. De flesta av dessa arbeten är dock begränsade till datorseende. Inom tidsseriernas område är detta relativt outforskat. I den här avhandlingen undersöks metoder för att använda tidsseriedata från en domän för att klassificera data från en annan domän med hjälp av djupinlärning. Fokus ligger på att använda accelerometerdata från löpning för att förbättra klassificeringen av hoppdata, baserat på den uppenbara likheten mellan löpning och hoppning. Således användes tekniker för överföringsinlärning och domänanpassning för att använda den inlärning som förvärvats genom träning av djupa modeller på löpsekvenser. I den här avhandlingen har fyra experiment utförts för att testa denna domänlikhet. Det första består av att omvandla tidsserier med den kontinuerliga wavelettransformen för att få fram både tids- och frekvensinformation. Modellen förtränas sedan inom en ram för kontrastiv inlärning. Användningen av CWT förbättrade dock inte klassificeringsresultaten. De följande två experimenten bestod av att förträna modellerna med självövervakad inlärning. Det första försöket med en kontrasterande förtextuppgift förbättrade klassificeringsresultaten och motståndskraften mot dataförlust. Det andra försöket med en prognostiserande förtextuppgift förbättrade resultaten när alla data var tillgängliga, men var mycket känslig för dataförlust. Slutligen testades domänanpassningen och visade intressanta resultat i klassificeringsuppgiften. Även om några av de använda teknikerna inte visade någon förbättring, har förträning med hjälp av kontrastinlärning på löpande dataset visat sig ge stora förbättringar när det gäller klassificering av hoppdata.

Semantic segmentation is the task of predicting predefined class labels for each pixel in a given image. It is essential in autonomous driving, but also challenging because training accurate models requires large and diverse datasets, which are difficult to collect due to the high cost of annotating images at pixel-level. This raises interest in using synthetic images from simulators, which can be labelled automatically. However, models trained directly on synthetic data perform poorly in real-world scenarios due to the distributional misalignment between synthetic and real images (domain shift). This thesis explores the effectiveness of several techniques for alleviating this issue, employing Synscapes and Cityscapes as the synthetic and real datasets, respectively. Some of the tested methods exploit a few additional labelled real images (few-shot supervised domain adaptation), some have access to plentiful real images but not their associated labels (unsupervised domain adaptation), and others do not take advantage of any image or annotation from the real domain (domain generalisation). After extensive experiments and a thorough comparative study, this work shows the severity of the domain shift problem by revealing that a semantic segmentation model trained directly on the synthetic dataset scores a poor mean Intersection over Union (mIoU) of 33:5% when tested on the real dataset. This thesis also demonstrates that such performance can be boosted by 25:7% without accessing any annotations from the real domain and 17:3% without leveraging any information from the real domain. Nevertheless, these gains are still inferior to the 31:0% relative improvement achieved with as little as 25 supplementary labelled real images, which suggests that there is still room for improvement in the fields of unsupervised domain adaptation and domain generalisation. Future work efforts should focus on developing better algorithms and creating synthetic datasets with a greater diversity of shapes and textures in order to reduce the domain shift. / Semantisk segmentering är uppgiften att förutsäga fördefinierade klassetiketter för varje pixel i en given bild. Det är viktigt för autonom körning, men också utmanande eftersom utveckling av noggranna modeller kräver stora och varierade datamängder, som är svåra att samla in på grund av de höga kostnaderna för att märka bilder på pixelnivå. Detta väcker intresset att använda syntetiska bilder från simulatorer, som kan märkas automatiskt. Problemet är emellertid att modeller som tränats direkt på syntetiska data presterar dåligt i verkliga scenarier på grund av fördelningsfel mellan syntetiska och verkliga bilder (domänskift). Denna avhandling undersöker effektiviteten hos flera tekniker för att lindra detta problem, med Synscapes och Cityscapes som syntetiska respektive verkliga datamängder. Några av de testade metoderna utnyttjar några ytterligare märkta riktiga bilder (few-shot övervakad domänanpassning), vissa har tillgång till många riktiga bilder men inte deras associerade etiketter (oövervakad domänanpassning), och andra drar inte nytta av någon bild eller annotering från den verkliga domänen (domängeneralisering). Efter omfattande experiment och en grundlig jämförande studie visar detta arbete svårighetsgraden av domänskiftproblemet genom att avslöja att en semantisk segmenteringsmodell som upplärts direkt på den syntetiska datauppsättningen ger en dålig mean Intersection over Union (mIoU) på 33; 5% när den testas på den verkliga datamängden. Denna avhandling visar också att sådan prestanda kan ökas med 25; 7% utan att komma åt några annoteringar från den verkliga domänen och 17; 3% utan att utnyttja någon information från den verkliga domänen. Ändå är dessa vinster fortfarande sämre än den 31; 0% relativa förbättringen som uppnåtts med så lite som 25 kompletterande annoterade riktiga bilder, vilket tyder på att det fortfarande finns utrymme för förbättringar inom områdena oövervakad domänanpassning och domängeneralisering. Framtida arbetsinsatser bör fokusera på att utveckla bättre algoritmer och på att skapa syntetiska datamängder med en större mångfald av former och texturer för att minska domänskiftet.

<p>  </p> <p>Large-scale deep learning models have reached previously unattainable performance for various tasks. However, the ever-growing resource consumption of neural networks generates large carbon footprint, brings difficulty for academics to engage in research and stops emerging economies from enjoying growing Artificial Intelligence (AI) benefits. To further scale AI to bring more benefits, two major challenges need to be solved. Firstly, even though large-scale deep learning models achieved remarkable success, their performance is still not satisfactory when fine-tuning with only a handful of examples, thereby hindering widespread adoption in real-world applications where a large scale of labeled data is difficult to obtain. Secondly, current machine learning models are still mainly designed for tasks in closed environments where testing datasets are highly similar to training datasets. When the deployed datasets have distribution shift relative to collected training data, we generally observe degraded performance of developed models. How to build adaptable models becomes another critical challenge. To address those challenges, in this dissertation, we focus on two topics: few-shot learning and domain adaptation, where few-shot learning aims to learn tasks with limited labeled data and domain adaption address the discrepancy between training data and testing data. In Part 1, we show our few-shot learning studies. The proposed few-shot solutions are built upon large-scale language models with evolutionary explorations from improving supervision signals, incorporating unlabeled data and improving few-shot learning abilities with lightweight fine-tuning design to reduce deployment costs. In Part 2, domain adaptation studies are introduced. We develop a progressive series of domain adaption approaches to transfer knowledge across domains efficiently to handle distribution shifts, including capturing common patterns across domains, adaptation with weak supervision and adaption to thousands of domains with limited labeled data and unlabeled data. </p>

No description available.

Machine learning systems have been demonstrated to be highly effective in various fields, such as in vision tasks for autonomous driving. However, the deployment of these systems poses a significant challenge in terms of ensuring their reliability and safety in diverse and dynamic environments. Online Unsupervised Domain Adaptation (UDA) aims to address the issue of continuous domain changes that may occur during deployment, such as sudden weather changes. Although these methods possess a remarkable ability to adapt to unseen domains, they are hindered by the high computational cost associated with constant adaptation, making them unsuitable for real-world applications that demand real-time performance. In this work, we focus on the challenging task of semantic segmentation. We present a framework for real-time domain adaptation that utilizes novel strategies to enable online adaptation at a rate of over 29 FPS on a single GPU. We propose a clever partial backpropagation in conjunction with a lightweight domain-shift detector that identifies the need for adaptation, adapting appropriately domain-specific hyperparameters to enhance performance. To validate our proposed framework, we conduct experiments in various storm scenarios using different rain intensities and evaluate our results in different domain shifts, such as fog visibility, and using the SHIFT dataset. Our results demonstrate that our framework achieves an optimal trade-off between accuracy and speed, surpassing state-of-the-art results, while the introduced strategies enable it to run more than six times faster at a minimal performance loss. / Maskininlärningssystem har visat sig vara mycket effektiva inom olika områden, till exempel i datorseende uppgifter för autonom körning. Spridning av dessa system utgör dock en betydande utmaning när det gäller att säkerställa deras tillförlitlighet och säkerhet i olika och dynamiska miljöer. Online Unsupervised Domain Adaptation (UDA) syftar till att behandla problemet med kontinuerliga domänändringar som kan inträffas under systemets användning, till exempel plötsliga väderförändringar. Även om dessa metoder har en anmärkningsvärd förmåga att anpassa sig till okända domäner, hindras de av den höga beräkningskostnaden som är förknippad med ständig nöndvändighet för anpassning, vilket gör dem olämpliga för verkliga tillämpningar som kräver realtidsprestanda. I detta avhandling fokuserar vi på utmanande uppgiften semantisk segmentering. Vi presenterar ett system för domänanpassning i realtid som använder nya strategier för att möjliggöra onlineanpassning med en hastighet av över 29 FPS på en enda GPU. Vi föreslår en smart partiell backpropagation i kombination med en lätt domänförskjutningsdetektor som identifierar nãr anpassning egentligen behövs, vilket kan konfigureras av domänspecifika hyperparametrar på lämpligt sätt för att förbättra prestandan. För att validera vårt föreslagna system genomför vi experiment i olika stormscenarier med olika regnintensiteter och utvärderar våra resultat i olika domänförskjutningar, såsom dimmasynlighet, och med hjälp av SHIFT-datauppsättningen. Våra resultat visar att vårt system uppnår en optimal avvägning mellan noggrannhet och hastighet, och överträffar toppmoderna resultat, medan de introducerade strategierna gör det möjligt att köra mer än sex gånger snabbare med minimal prestandaförlust.

<p>Domain adaptation is one of the hottest directions in solving annotation insufficiency problem of deep learning. General domain adaptation is not consistent with the practical scenarios in the industry. In this thesis, we focus on two concerns as below.</p> <p>  </p> <p>  First is that labeled data are generally collected from multiple domains. In other words, multi-source adaptation is a more common situation. Simply extending these single-source approaches to the multi-source cases could cause sub-optimal inference, so specialized multi-source adaptation methods are essential. The main challenge in the multi-source scenario is a more complex divergence situation. Not only the divergence between target and each source plays a role, but the divergences among distinct sources matter as well. However, the significance of maintaining consistency among multiple sources didn't gain enough attention in previous work. In this thesis, we propose an Enhanced Consistency Multi-Source Adaptation (EC-MSA) framework to address it from three perspectives. First, we mitigate feature-level discrepancy by cross-domain conditional alignment, narrowing the divergence between each source and target domain class-wisely. Second, we enhance multi-source consistency via dual mix-up, diminishing the disagreements among different sources. Third, we deploy a target distilling mechanism to handle the uncertainty of target prediction, aiming to provide high-quality pseudo-labeled target samples to benefit the previous two aspects. Extensive experiments are conducted on several common benchmark datasets and demonstrate that our model outperforms the state-of-the-art methods.</p> <p>  </p> <p>  Second is that data privacy and security is necessary in practice. That is, we hope to keep the raw data stored locally while can still obtain a satisfied model. In such a case, the risk of data leakage greatly decreases. Therefore, it is natural for us to combine the federated learning paradigm with domain adaptation. Under the source-private setting, the main challenge for us is to expose information from the source domain to the target domain while make sure that the communication process is safe enough. In this thesis, we propose a method named Fourier Transform-Assisted Federated Domain Adaptation (FTA-FDA) to alleviate the difficulties in two ways. We apply Fast Fourier Transform to the raw data and transfer only the amplitude spectra during the communication. Then frequency space interpolations between these two domains are conducted, minimizing the discrepancies while ensuring the contact of them and keeping raw data safe. What's more, we make prototype alignments by using the model weights together with target features, trying to reduce the discrepancy in the class level. Experiments on Office-31 demonstrate the effectiveness and competitiveness of our approach, and further analyses prove that our algorithm can help protect privacy and security.</p>

[pt] Os dados de observação da Terra são freqüentemente afetados pelo fenômeno de mudança de domínio. Mudanças nas condições ambientais, variabilidade geográfica e diferentes propriedades de sensores geralmente tornam quase impossível empregar classificadores previamente treinados para novos dados sem experimentar uma queda significativa na precisão da classificação. As técnicas de adaptação de domínio baseadas em modelos de aprendizado profundo têm se mostrado úteis para aliviar o problema da mudança de domínio. Trabalhos recentes nesta área fundamentam-se no treinamento adversárial para alinhar os atributos extraídos de imagens de diferentes domínios em um espaço latente comum. Outra forma de tratar o problema é empregar técnicas de translação de imagens e adaptá-las de um domínio para outro de forma que as imagens transformadas contenham características semelhantes às imagens do outro domínio. Neste trabalho, propõem-se abordagens de adaptação de domínio para tarefas de detecção de mudanças, baseadas em primeiro lugar numa técnica de traslação de imagens, Cycle-Consistent Generative Adversarial Network (CycleGAN), e em segundo lugar, num modelo de alinhamento de atributos: a Domain Adversarial Neural Network (DANN). Particularmente, tais técnicas foram estendidas, introduzindo-se restrições adicionais na fase de treinamento dos componentes do modelo CycleGAN, bem como um procedimento de pseudo-rotulagem não supervisionado para mitigar o impacto negativo do desequilíbrio de classes no DANN. As abordagens propostas foram avaliadas numa aplicação de detecção de desmatamento, considerando diferentes regiões na floresta amazônica e no Cerrado brasileiro (savana). Nos experimentos, cada região corresponde a um domínio, e a precisão de um classificador treinado com imagens e referências de um dos domínio (fonte) é medida na classificação de outro domínio (destino). Os resultados demonstram que as abordagens propostas foram bem sucedidas em amenizar o problema de desvio de domínio no contexto da aplicação alvo. / [en] Earth observation data are frequently affected by the domain shift phenomenon. Changes in environmental conditions, geographical variability and different sensor properties typically make it almost impossible to employ previously trained classifiers for new data without a significant drop in classification accuracy. Domain adaptation (DA) techniques based on Deep Learning models have been proven useful to alleviate domain shift. Recent improvements in DA technology rely on adversarial training to align features extracted from images of the different domains in a common latent space. Another way to face the problem is to employ image translation techniques, and adapt images from one domain in such a way that the transformed images contain characteristics that are similar to the images from the other domain. In this work two different DA approaches for change detection tasks are proposed, which are based on a particular image translation technique, the Cycle-Consistent Generative Adversarial Network (CycleGAN), and on a representation matching strategy, the Domain Adversarial Neural Network (DANN). In particular, additional constraints in the training phase of the original CycleGAN model components are proposed, as well as an unsupervised pseudo-labeling procedure, to mitigate the negative impact of class imbalance in the DANN-based approach. The proposed approaches were evaluated on a deforestation detection application, considering different sites in the Amazon rain-forest and in the Brazilian Cerrado (savanna) biomes. In the experiments each site corresponds to a domain, and the accuracy of a classifier trained with images and references from one (source) domain is measured in the classification of another (target) domain. The experimental results show that the proposed approaches are successful in alleviating the domain shift problem.