We introduce DIP, a novel unsupervised post-training method designed to enhance dense image representations in large-scale pretrained vision encoders for in-context scene understanding. Unlike prior approaches that rely on complex self-distillation architectures, our method trains the vision encoder using pseudo-tasks that explicitly simulate downstream in-context scenarios, inspired by meta-learning principles. To enable post-training on unlabeled data, we propose an automatic mechanism for generating in-context tasks that combines a pretrained diffusion model and the vision encoder itself. DIP is simple, unsupervised, and computationally efficient, requiring less than 9 hours on a single A100 GPU. By learning dense representations through pseudo in-context tasks, it achieves strong performance across a wide variety of downstream real-world in-context scene understanding tasks. It outperforms both the initial vision encoder and prior methods, offering a practical and effective solution for improving dense representations. Code available here: https://github.com/sirkosophia/DIP

Abstract Semantic image segmentation models typically require extensive pixel-wise annotations, which are costly to obtain and prone to biases. Our work investigates learning semantic segmentation in urban scenes without any manual annotation. We propose a novel method for learning pixel-wise semantic segmentation using raw, uncurated data from vehicle-mounted cameras and LiDAR sensors, thus eliminating the need for manual labeling. Our contributions are as follows. First, we develop a novel approach for cross-modal unsupervised learning of semantic segmentation by leveraging synchronized LiDAR and image data. A crucial element of our method is the integration of an object proposal module that examines the LiDAR point cloud to generate proposals for spatially consistent objects. Second, we demonstrate that these 3D object proposals can be aligned with corresponding images and effectively grouped into semantically meaningful pseudo-classes. Third, we introduce a cross-modal distillation technique that utilizes image data partially annotated with the learnt pseudo-classes to train a transformer-based model for semantic image segmentation. Fourth, we demonstrate further significant improvements of our approach by extending the proposed model using a teacher-student distillation with an exponential moving average and incorporating soft targets from the teacher. We show the generalization capabilities of our method by testing on four different testing datasets (Cityscapes, Dark Zurich, Nighttime Driving, and ACDC) without any fine-tuning. We present an in-depth experimental analysis of the proposed model including results when using another pre-training dataset, per-class and pixel accuracy results, confusion matrices, PCA visualization, k-NN evaluation, ablations of the number of clusters and LiDAR’s density, supervised finetuning as well as additional qualitative results and their analysis.

Recent advances in self-supervised visual representation learning have paved the way for unsupervised methods tackling tasks such as object discovery and instance segmentation. However, discovering objects in an image with no supervision is a very hard task; what are the desired objects, when to separate them into parts, how many are there, and of what classes? The answers to these questions de-pend on the tasks and datasets of evaluation. In this work, we take a different approach and propose to look for the background instead. This way, the salient objects emerge as a by-product without any strong assumption on what an object should be. We propose FOUND, a simple model made of a single conv1 x 1 initialized with coarse background masks extracted from self-supervised patch-based representations. After fast training and refining these seed masks, the model reaches state-of-the-art results on unsupervised saliency detection and object discovery benchmarks. Moreover, we show that our approach yields good results in the unsupervised semantic segmentation retrieval task. The code to reproduce our results is available at https://github.com/valeoai/FOUND.

We describe an approach to predict open-vocabulary 3D semantic voxel occupancy map from input 2D images with the objective of enabling 3D grounding, segmentation and retrieval of free-form language queries. This is a challenging problem because of the 2D-3D ambiguity and the open-vocabulary nature of the target tasks, where obtaining annotated training data in 3D is difficult. The contributions of this work are three-fold. First, we design a new model architecture for open-vocabulary 3D semantic occupancy prediction. The architecture consists of a 2D-3D encoder together with occupancy prediction and 3D-language heads. The output is a dense voxel map of 3D grounded language embeddings enabling a range of open-vocabulary tasks. Second, we develop a tri-modal self-supervised learning algorithm that leverages three modalities: (i) images, (ii) language and (iii) LiDAR point clouds, and enables training the proposed architecture using a strong pre-trained vision-language model without the need for any 3D manual language annotations. Finally, we demonstrate quantitatively the strengths of the proposed model on several open-vocabulary tasks: Zero-shot 3D semantic segmentation using existing datasets; 3D grounding and retrieval of free-form language queries, using a small dataset that we propose as an extension of nuScenes. You can find the project page here https://vobecant.github.io/POP3D.

Existing datasets for training pedestrian detectors in images suffer from limited appearance and pose variation. The most challenging scenarios are rarely included because they are too difficult to capture due to safety reasons, or they are very unlikely to happen. The strict safety requirements in assisted and autonomous driving applications call for an extra high detection accuracy also in these rare situations. Having the ability to generate people images in arbitrary poses, with arbitrary appearances and embedded in different background scenes with varying illumination and weather conditions, is a crucial component for the development and testing of such applications. The contributions of this paper are three-fold. First, we describe an augmentation method for the controlled synthesis of urban scenes containing people, thus producing rare or never-seen situations. This is achieved with a data generator (called DummyNet) with disentangled control of the pose, the appearance, and the target background scene. Second, the proposed generator relies on novel network architecture and associated loss that takes into account the segmentation of the foreground person and its composition into the background scene. Finally, we demonstrate that the data generated by our DummyNet improve the performance of several existing person detectors across various datasets as well as in challenging situations, such as night-time conditions, where only a limited amount of training data is available. In the setup with only day-time data available, we improve the night-time detector by 17% log-average miss rate over the detector trained with the day-time data only.

Wide-angle fisheye cameras are commonly used in automated driving for parking and low-speed navigation tasks. Four of such cameras form a surround-view system that provides a complete and detailed view of the vehicle. These cameras are directly exposed to harsh environmental settings and can get soiled very easily by mud, dust, water, frost. Soiling on the camera lens can severely degrade the visual perception algorithms, and a camera cleaning system triggered by a soiling detection algorithm is increasingly being deployed. While adverse weather conditions, such as rain, are getting attention recently, there is only limited work on general soiling. The main reason is the difficulty in collecting a diverse dataset as it is a relatively rare event.We propose a novel GAN based algorithm for generating unseen patterns of soiled images. Additionally, the proposed method automatically provides the corresponding soiling masks eliminating the manual annotation cost. Augmentation of the generated soiled images for training improves the accuracy of soiling detection tasks significantly by 18% demonstrating its usefulness. The manually annotated soiling dataset and the generated augmentation dataset will be made public. We demonstrate the generalization of our fisheye trained GAN model on the Cityscapes dataset. We provide an empirical evaluation of the degradation of the semantic segmentation algorithm with the soiled data.

In situations when potentially costly decisions are being made, faces of people tend to reflect a level of certainty about the appropriateness of the chosen decision. This fact is known from the psychological literature. In the paper, we propose a method that uses facial images for automatic detection of the decision ambiguity state of a subject. To train and test the method, we collected a large-scale dataset from "Who Wants to Be a Millionaire?" -- a popular TV game show. The videos provide examples of various mental states of contestants, including uncertainty, doubts and hesitation. The annotation of the videos is done automatically from on-screen graphics. The problem of detecting decision ambiguity is formulated as binary classification. Video-clips where a contestant asks for help (audience, friend, 50:50) are considered as positive samples; if he (she) replies directly as negative ones. We propose a baseline method combining a deep convolutional neural network with an SVM. The method has an error rate of 24%. The error of human volunteers on the same dataset is 45%, close to chance.