Umbo Computer Vision (United Kingdom)

Single image scene relighting aims to generate a realistic new version of an input image so that it appears to be illuminated by a new target light condition. Although existing works have explored this problem from various perspectives, generating relit images under arbitrary light conditions remains highly challenging, and related datasets are scarce. Our work addresses this problem from both the dataset and methodological perspectives. We propose two new datasets: a synthetic dataset with the ground truth of intrinsic components and a real dataset collected under laboratory conditions. These datasets alleviate the scarcity of existing datasets. To incorporate physical consistency in the relighting pipeline, we establish a two-stage network based on intrinsic decomposition, giving outputs at intermediate steps, thereby introducing physical constraints. When the training set lacks ground truth for intrinsic decomposition, we introduce an unsupervised module to ensure that the intrinsic outputs are satisfactory. Our method outperforms the state-of-the-art methods in performance, as tested on both existing datasets and our newly developed datasets. Furthermore, pretraining our method or other prior methods using our synthetic dataset can enhance their performance on other datasets. Since our method can accommodate any light conditions, it is capable of producing animated results.

We present a virtual model to optimize point of entry (POE) in lung biopsy planning systems. Our model allows to compute the quality of a biopsy sample taken from potential POE. Biopsy quality is given by a heatmap projected onto the skeleton of a patient-specific model of airways. The skeleton provides a 3D representation of airways structure, while the heatmap intensity represents the maximum amount of tissue that could be extracted from each POE. This amount of tissue is determined by the intersection of the lesion with a cone that represents an uncertainty area in the introduction of biopsy instruments. The cone, lesion, and skeleton define a 3D scene. The analysis of the heatmaps for different settings of the 3 elements of the scene allow to draw conclusions about the impact on the biopsy of lesion characteristics, like shape, orientation and distance to bronchi.

This paper highlights the need to bring document classification benchmarking closer to real-world applications, both in the nature of data tested (X: multi-channel, multipaged, multi-industry; Y : class distributions and label set variety) and in classification tasks considered (f : multipage document, page stream, and document bundle classification, ...). We identify the lack of public multi-page document classification datasets, formalize different classification tasks arising in application scenarios, and motivate the value of targeting efficient multi-page document representations. An experimental study on proposed multi-page document classification datasets demonstrates that current benchmarks have become irrelevant and need to be updated to evaluate complete documents, as they naturally occur in practice. This reality check also calls for more mature evaluation methodologies, covering calibration evaluation, inference complexity (time-memory), and a range of realistic distribution shifts (e.g., born-digital vs. scanning noise, shifting page order). Our study ends on a hopeful note by recommending concrete avenues for future improvements.

Deployments of real-world object-detection systems often experience a degradation in performance over time due to concept drift. Systems that leverage thermal cameras are especially susceptible because the respective thermal signatures of objects and their surroundings are highly sensitive to environmental changes. In this study, a conditioning method is investigated. The method aims to guide the training loop of thermal object detection systems by leveraging an auxiliary branch to predict the weather, while directly or indirectly conditioning the baseline detection system. Leveraging such an approach to train detection networks does not necessarily improve the performance of native architectures, however, it can be observed that conditioned networks manage to extract a signal from thermal images that guides the network to detect objects that baseline models miss. As the extracted signal appears to be quite noisy and very challenging to regress accurately, further work is needed to identify an ideal optimization vector.

Despite decades of research in Optical Music Recognition (OMR), the recognition of old handwritten music scores remains a challenge because of the variabilities in the handwriting styles, paper degradation, lack of standard notation, etc. Therefore, the research in OMR systems adapted to the particularities of old manuscripts is crucial to accelerate the conversion of music scores existing in archives into digital libraries, fostering the dissemination and preservation of our music heritage. In this paper we explore the adaptation of sequence-to-sequence models with attention mechanism (used in translation and handwritten text recognition) and the generation of specific synthetic data for recognizing old music scores. The experimental validation demonstrates that our approach is promising, especially when compared with long short-term memory neural networks.

Person re-identification (re-ID) aims at matching images of the same identity across camera views. Due to varying distances between cameras and persons of interest, resolution mismatch can be expected, which would degrade person re-ID performance in real-world scenarios. To overcome this problem, we propose a novel generative adversarial network to address cross-resolution person re-ID, allowing query images with varying resolutions. By advancing adversarial learning techniques, our proposed model learns resolution-invariant image representations while being able to recover the missing details in low-resolution input images. The resulting features can be jointly applied for improving person re-ID performance due to preserving resolution invariance and recovering re-ID oriented discriminative details. Our experiments on five benchmark datasets confirm the effectiveness of our approach and its superiority over the state-of-the-art methods, especially when the input resolutions are unseen during training.

Traffic sign classification is a challenging problem in Computer Vision due to the high variability of sign appearance in uncontrolled environments. Lack of visibility, illumination changes, and partial occlusions are just a few problems. In this paper, we introduce a classification technique for traffic signs recognition by means of Error Correcting Output Codes. Recently, new proposals of coding and decoding strategies for the Error Correcting Output Codes framework have been shown to be very effective in front of multiclass problems. We review the state-of-the-art ECOC strategies and combinations of problem-dependent coding designs and decoding techniques. We apply these approaches to the Mobile Mapping problem. We detect the sign regions by means of Adaboost. The Adaboost in an attentional cascade with the extended set of Haar-like features estimated on the integral shows great performance at the detection step. Then, a spatial normalization using the Hough transform and the fast radial symmetry is done. The model fitting improves the final classification performance by normalizing the sign content. Finally, we classify a wide set of traffic signs types, obtaining high success in adverse conditions.

We present an update of the lunar occultation program which is routinely carried out in the near-IR at the Calar Alto Observatory. A total of 350 events were recorded since our last report (Fors et al. 2004). In the course of eight runs we have observed, among others, late-type giants, T-Tauri stars, and infrared sources. Noteworthy was a passage of the Moon close to the galactic center, which produced a large number of events during just a few hours in July 2004. Results include the determinations of the angular diameter of RZ Ari, and the projected separations and brightness ratios for one triple and 13 binary stars, almost all of which representing first time detections. Projected separations range from 0.09arcsec to 0.007arcsec. We provide a quantitative analysis of the performance achieved in our observations in terms of angular resolution and sensitivity, which reach about 0.003arcsec and K~8.5mag, respectively. We also present a statistical discussion of our sample, and in particular of the frequency of detection of binaries among field stars.