Diffusion models have emerged as highly effective techniques for inpainting, however, they remain constrained by slow sampling rates. While recent advances have enhanced generation quality, they have also increased sampling time, thereby limiting scalability in real-world applications. We investigate the generative sampling process of diffusion-based inpainting models and observe that these models make minimal use of the input condition during the initial sampling steps. As a result, the sampling trajectory deviates from the data manifold, requiring complex synchronization mechanisms to realign the generation process. To address this, we propose Time-aware Diffusion Paint (TD-Paint), a novel approach that adapts the diffusion process by modeling variable noise levels at the pixel level. This technique allows the model to efficiently use known pixel values from the start, guiding the generation process toward the target manifold. By embedding this information early in the diffusion process, TD-Paint significantly accelerates sampling without compromising image quality. Unlike conventional diffusion-based inpainting models, which require a dedicated architecture or an expensive generation loop, TD-Paint achieves faster sampling times without architectural modifications. Experimental results across three datasets show that TD-Paint outperforms state-of-the-art diffusion models while maintaining lower complexity.

In this paper, we study the problem of multi-domain translation: given an element (a) of domain A, we wish to generate a corresponding element (b) in another domain B, and vice versa. Acquiring supervision in multiple domains can be a tedious task, also we propose to learn this translation from one domain to another when supervision is available as a pair (a, b), and leverage possible unpaired data when only (a) or only (b) is available. We introduce a new unified framework called Latent Space Mapping (LSM), which exploits the manifold assumption to learn a latent space from each domain. Unlike existing approaches, we propose to further regularize each latent space using available domains by learning each dependency between pairs of domains. We evaluate our approach on three tasks performing i) a synthetic dataset with image translation, ii) a real-world task of semantic segmentation for medical images, and iii) a real-world task of facial landmark detection.

Contrastive representation learning has proven to be an effective self-supervised learning method. Most successful approaches are based on Noise Contrastive Estimation (NCE) and use different views of an instance as positives that should be contrasted with other instances, called negatives, that are considered as noise. However, several instances in a dataset are drawn from the same distribution and share underlying semantic information. A good data representation should contain relations, or semantic similarity, between the instances. Contrastive learning implicitly learns relations but considering all negatives as noise harms the quality of the learned relations. To circumvent this issue, we propose a novel formulation of contrastive learning using semantic similarity between instances called Similarity Contrastive Estimation (SCE). Our training objective is a soft contrastive learning one. Instead of hard classifying positives and negatives, we estimate from one view of a batch a continuous distribution to push or pull instances based on their semantic similarities. This target similarity distribution is sharpened to eliminate noisy relations. The model predicts for each instance, from another view, the target distribution while contrasting its positive with negatives. Experimental results show that SCE is Top-1 on the ImageNet linear evaluation protocol at 100 pretraining epochs with 72.1% accuracy and is competitive with state-of-the-art algorithms by reaching 75.4% for 200 epochs with multi-crop. We also show that SCE is able to generalize to several tasks. Source code is available here: https://github.com/CEA-LIST/SCE.

Abstract Contrastive representation learning has proven to be an effective self-supervised learning method for images and videos. Most successful approaches are based on Noise Contrastive Estimation (NCE) and use different views of an instance as positives that should be contrasted with other instances, called negatives, that are considered as noise. However, several instances in a dataset are drawn from the same distribution and share underlying semantic information. A good data representation should contain relations between the instances, or semantic similarity and dissimilarity, that contrastive learning harms by considering all negatives as noise. To circumvent this issue, we propose a novel formulation of contrastive learning using semantic similarity between instances called Similarity Contrastive Estimation (SCE). Our training objective is a soft contrastive one that brings the positives closer and estimates a continuous distribution to push or pull negative instances based on their learned similarities. We validate empirically our approach on both image and video representation learning. We show that SCE performs competitively with the state of the art on the ImageNet linear evaluation protocol for fewer pretraining epochs and that it generalizes to several downstream image tasks. We also show that SCE reaches state-of-the-art results for pretraining video representation and that the learned representation can generalize to video downstream tasks. Source code is available here: https://github.com/juliendenize/eztorch .

Chord represents the back-bone of occidental music genre as it contains rich harmonic information which is useful for various music applications such as music genre classification or music retrieval. Hence, chord recognition or transcription is of importance for music representation. In this paper we focus on chord recognition and especially investigate different features representation used in such a system: classical features as well as a new type of feature we propose are explored. We evaluate their usefulness through a multi-class chord classification problem.

State-of-the-art pattern recognition methods have difficulties dealing with problems where the dimension of the output space is large. In this article, we propose a framework based on deep architectures (e. g. Deep Neural Networks) in order to deal with this issue. Deep architectures have proven to be efficient for high dimensional input problems such as image classification, due to their ability to embed the input space. The main contribution of this article is the extension of the embedding procedure to both the input and output spaces to easily handle complex outputs. Using this extension, inter-output dependencies can be modelled efficiently. This provides an interesting alternative to probabilistic models such as HMM and CRF. Preliminary experiments on toy datasets and USPS character reconstruction show promising results.

The scores returned by support vector machines are often used as a confidence measures in the classification of new examples. However, there is no theoretical argument sustaining this practice. Thus, when classification uncertainty has to be assessed, it is safer to resort to classifiers estimating conditional probabilities of class labels. Here, we focus on the ambiguity in the vicinity of the boundary decision. We propose an adaptation of maximum likelihood estimation, instantiated on logistic regression. The model outputs proper conditional probabilities into a user-defined interval and is less precise elsewhere. The model is also sparse, in the sense that few examples contribute to the solution. The computational efficiency is thus improved compared to logistic regression. Furthermore, preliminary experiments show improvements over standard logistic regression and performances similar to support vector machines.