Our paper on the use of Kalman filtering for neural networks optimization got accepted to AAAI2022!
Aram Davtyan, Sepehr Sameni, Llukman Cerkezi, Givi Meishvili, Adam Bielski and Paolo Favaro, in AAAI Conference on Artificial Intelligence, 2022.
Optimization is often cast as a deterministic problem, where the solution is found through some iterative procedure such as gradient descent. However, when training neural networks the loss function changes over (iteration) time due to the randomized selection of a subset of the samples. This randomization turns the optimization problem into a stochastic one. We propose to consider the loss as a noisy observation with respect to some reference optimum. This interpretation of the loss allows us to adopt Kalman filtering as an optimizer, as its recursive formulation is designed to estimate unknown parameters from noisy measurements. Moreover, we show that the Kalman Filter dynamical model for the evolution of the unknown parameters can be used to capture the gradient dynamics of advanced methods such as Momentum and Adam. We call this stochastic optimization method KOALA, which is short for Kalman Optimization Algorithm with Loss Adaptivity. KOALA is an easy to implement, scalable, and efficient method to train neural networks. We provide convergence analysis and show experimentally that it yields parameter estimates that are on par with or better than existing state of the art optimization algorithms across several neural network architectures and machine learning tasks, such as computer vision and language modeling.
Paper: https://arxiv.org/pdf/2107.03331.pdf
Multiple Master thesis projects are offered at the ARTORG Center and Prophesee.
The Gerontechnology and Rehabilitation group at the ARTORG Center for Biomedical Engineering is offering multiple MSc thesis projects to students, which are interested in working with real patient data, artificial intelligence and machine learning algorithms. The goal of these projects is to transfer the findings to the clinic in order to solve today’s healthcare problems and thus to improve the quality of life of patients.
- Assessment of Digital Biomarkers at Home by Radar. [PDF]
- Comparison of Radar, Seismograph and Ballistocardiography and to Monitor Sleep at Home. [PDF]
- Sentimental Analysis in Speech. [PDF]
Contact: Dr. Stephan Gerber (stephan.gerber@artorg.unibe.ch
)
A 6 month intership at Prophesee, Grenoble is offered to a talented Master Student.
The topic of the internship is working on burst imaging following the work of Sam Hasinoff, and exploring ways to improve it using event-based vision.
A compensation to cover the expenses of living in Grenoble is offered. Only students that have legal rights to work in France can apply.
Anyone interested can send an email with the CV to Daniele Perrone (dperrone@prophesee.ai
More thesis topics from the Computer Vision Group can be found here.
We have new research papers published in top ML and CV conferences!
Givi Meishvili, Attila Szabo, Simon Jenni and Paolo Favaro, in British Machine Vision Conference (BMVC), 2021.
We propose a solution to the novel task of rendering sharp videos from new view-points from a single motion-blurred image of a face. Our method handles the complexity of face blur by implicitly learning the geometry and motion of faces through the joint training on three large datasets: FFHQ and 300VW, which are publicly available, and a new multi-view face dataset that we built, which will be made available upon publication. The first two datasets provide a large variety of faces and allow our model to generalize better. The third dataset instead allows us to introduce multi-view constraints, which are crucial to synthesizing sharp videos from a new camera view. Our dataset consists of high frame rate synchronized videos from multiple views of several subjects displaying a wide range of facial expressions. We use the high frame rate videos to simulate real-istic motion blur through averaging. Thanks to this dataset, we train a neural network to reconstruct a 3D video representation from a single image and the corresponding face gaze. We then provide a camera viewpoint relative to the estimated gaze and the blurry image as input to an encoder-decoder network to generate a video of sharp frames with anovel camera viewpoint. We demonstrate our approach on test subjects of our multi-view dataset and VIDTIMIT.
Paper: https://www.bmvc2021-virtualconference.com/assets/papers/1043.pdf
Ajinkya Tejankar*, Soroush Abbasi Koohpayegani*, Vipin Pillai, Paolo Favaro, Hamed Pirsiavash, in International Conference on Computer Vision (ICCV), 2021.
Recently, contrastive learning has achieved great results in self-supervised learning, where the main idea is to pull two augmentations of an image (positive pairs) closer compared to other random images (negative pairs). We argue that not all negative images are equally negative. Hence, we introduce a self-supervised learning algorithm where we use a soft similarity for the negative images rather than a binary distinction between positive and negative pairs. We iteratively distill a slowly evolving teacher model to the student model by capturing the similarity of a query image to some random images and transferring that knowledge to the student. Specifically, our method should handle unbalanced and unlabeled data better than existing contrastive learning methods, because the randomly chosen negative set might include many samples that are semantically similar to the query image. In this case, our method labels them as highly similar while standard contrastive methods label them as negatives. Our method achieves comparable results to the state-of-the-art models. Our code is available here: https://github.com/UMBCvision/ISD.
Paper: https://www.cvg.unibe.ch/media/publications/pdf/ISD_iccv21.pdf
Tomoki Watanabe and Paolo Favaro, in International Conference on Machine Learning (ICML), 2021.
We propose a novel GAN training scheme that can handle any level of labeling in a unified manner. Our scheme introduces a form of artificial labeling that can incorporate manually defined labels, when available, and induce an alignment between them. To define the artificial labels, we exploit the assumption that neural network generators can be trained more easily to map nearby latent vectors to data with semantic similarities, than across separate categories. We use generated data samples and their corresponding artificial conditioning labels to train a classifier. The classifier is then used to self-label real data. To boost the accuracy of the self-labeling, we also use the exponential moving average of the classifier. However, because the classifier might still make mistakes, especially at the beginning of the training, we also refine the labels through self-attention, by using the labeling of real data samples only when the classifier outputs a high classification probability score. We evaluate our approach on CIFAR-10, STL-10 and SVHN, and show that both self-labeling and self-attention consistently improve the quality of generated data. More surprisingly, we find that the proposed scheme can even outperform classconditional GANs.
Paper: https://arxiv.org/pdf/2106.09914.pdf
The Gerontechnology and Rehabilitation group at the ARTORG Center for Biomedical Engineering is offering multiple BSc- and MSc thesis projects to students, which are interested in working with real patient data, artificial intelligence and machine learning algorithms. The goal of these projects is to transfer the findings to the clinic in order to solve today’s healthcare problems and thus to improve the quality of life of patients.
Contact: Dr. Stephan Gerber (stefan.gerber@artorg.unibe.ch
More thesis topics from the Computer Vision Group can be found here.
The pandemic can't stop us! We have two new research papers published in GCPR and ACCV. Both were selected to be oral presentations.
Abdelhak Lemkhenter and Paolo Favaro, in German Conference on Pattern Recognition (GCPR), 2020.
In this work, we introduce Phase-Swap, a novel self-supervised learning task for bio-signals. Most hand-crafted features for bio-signals in general, and EEG in particular, are derived from the power-spetrum, e.g. considering the energy of the signal within predefined frequency bands. In fact, most often the phase information is discared as it is more sample specific, and thus more sensitive to noise, than the amplitude. However, various medical studies have shown the link between the phase component and various physiological patterns such as cognitive functions in the case of brain activity. Motivated by this line of research, we build a self-supervised task that encourages the trained models to learn the implicit phase-amplitude coupling. This task, named Phase Swap, consists of discriminating between real samples and samples for which the phase component in the fourrier domain was swapped out by one taken from another sample. We show that the learned self-supervised features generalize better across experimental settings and subject identities compared to a supervised baseline for two classification tasks, seizure
detection and sleep scoring, on four different dataset: ExpandedEDF (Sleep Cassette + Sleep Telemetry), CHB-MIT and the ISRUC-Sleep data set. These findings highlight the benefits our self-supervised pretraining for various machine learning applications for bio-signals.
Pre-print: https://arxiv.org/pdf/2009.07664
Simon Jenni and Paolo Favaro, in Asian Conference on Computer Vision (ACCV), 2020.
Current state of the art methods cast monocular 3D human pose estimation as a learning problem by training neural networks on costly large data sets of images and corresponding skeleton poses. In contrast, we propose an approach that can exploit small annotated data sets by fine-tuning networks pre-trained via self-supervised learning on (large) unlabeled data sets. To drive such models in the pre-training step towards supporting 3D pose estimation, we introduce a novel self-supervised feature learning task designed to focus on the 3D structure in an image. We exploit images extracted from videos captured with a multi-view camera system. The task is to classify whether two images depict two views of the same scene up to a rigid transformation. In a multi-view data set, where objects deform in a non rigid manner, a rigid transformation occurs only between two views taken at the exact same time, i.e., when they are synchronized. We demonstrate the effectiveness of the synchronization task on the Human3.6M data set and achieve state of the art results in 3D human pose estimation.