Institute Talks

Building Multi-Family Animal Models

Talk
  • 07 April 2017 • 11:00 12:00
  • Silvia Zuffi
  • Aquarium, N.3.022, Spemannstr. 34, third floor

There has been significant prior work on learning realistic, articulated, 3D statistical shape models of the human body. In contrast, there are few such models for animals, despite their many applications in biology, neuroscience, agriculture, and entertainment. The main challenge is that animals are much less cooperative subjects than humans: the best human body models are learned from thousands of 3D scans of people in specific poses, which is infeasible with live animals. In the talk I will illustrate how we extend a state-of-the-art articulated 3D human body model (SMPL) to animals learning from toys a multi-family shape space that can represent lions, cats, dogs, horses, cows and hippos. The generalization of the model is illustrated by fitting it to images of real animals, where it captures realistic animal shapes, even for new species not seen in training.

Frederick Eberhardt - TBA

IS Colloquium
  • 03 July 2017 • 11:15 12:15
  • Frederick Eberhardt
  • Max Planck House Lecture Hall

Organizers: Sebastian Weichwald

  • Bojan Pepik

Current object class detection methods typically target 2D bounding box localization, encouraged by benchmark data sets, such as Pascal VOC. While this seems suitable for the detection of individual objects, higher-level applications, such as autonomous driving and 3D scene understanding, would benefit from more detailed and richer object hypotheses. In this talk I will present our recent work on building more detailed object class detectors, bridging the gap between higher level tasks and state-of-the-art object detectors. I will present a 3D object class detection method that can reliably estimate the 3D position, orientation and 3D shape of objects from a single image. Based on state-of-the-art CNN features, the method is a carefully designed 3D detection pipeline where each step is tuned for better performance, resulting in a registered CAD model for every object in the image. In the second part of the talk, I will focus on our work on what is holding back convolutional neural nets for detection. We analyze the R-CNN object detection pipeline in combination with state-of-the-art network architectures (AlexNet, GoogleNet and VGG16). Focusing on two central questions, what did the convnets learn and what can they learn, we illustrate that the three network architectures suffer from the same weaknesses, and these downsides can not be alleviated by simply introducing more data. Therefore we conclude that architectural changes are needed. Furthermore, we show that additional, synthetical generated training data, sampled from the modes of the data distribution can further increase the overall detection performance, while still suffering from the same weaknesses. Last, we hint at the complementary nature of the features of the three network architectures considered in this work.


Articulated motion discovery using pairs of trajectories

Talk
  • 26 August 2015 • 11:15 12:15
  • Luca del Pero
  • MRC Seminar Room, Spemannstr. 41

Most computer vision systems cannot take advantage of the abundance of Internet videos as training data. This is because current methods typically learn under strong supervision and require expensive manual annotations. (e.g. videos need to be temporally trimmed to cover the duration of a specific action, object bounding boxes, etc.). In this talk, I will present two techniques that can lead to learning the behavior and the structure of articulated object classes (e.g. animals) from videos, with as little human supervision as possible. First, we discover the characteristic motion patterns of an object class from videos of objects performing natural, unscripted behaviors, such as tigers in the wild. Our method generates temporal intervals that are automatically trimmed to one instance of the discovered behavior, and clusters them by type (e.g. running, turning head, drinking water). Second, we automatically recover thousands of spatiotemporal correspondences within the discovered clusters of behavior, which allow mapping pixels of an instance in one video to those of a different instance in a different video. Both techniques rely on a novel motion descriptor modeling the relative displacement of pairs of trajectories, which is more suitable for articulated objects than state-of-the-art descriptors using single trajectories. We provide extensive quantitative evaluation on our new dataset of tiger videos, which contains more than 100k fully annotated frames.

Organizers: Laura Sevilla


Novel optical sensing technologies

Talk
  • 29 July 2015 • 10:30 11:00
  • Felipe Guzman
  • AGBS Seminar Room

Coherent light enables optical measurements of exquisite sensitivity, advancing technologies for improved sensing and autonomous systems. From my previous work on gravitational physics, I will present a brief overview on technologies for laser-interferometric gravitational-wave observatories, particularly within the scope of the European mission LISA Pathfinder, to be launched at the end 2015. In addition, I will talk about my current work in novel and highly compact optomechanical systems and photonic crystals, optical micro-resonators with sensitivities below femtometer levels, as well as fiber-based non-destructive inspection techniques. To conclude, I will share my ideas on how to expand my research in optomechanical systems, optical technologies, and real-time data and image processing towards applications in robotics and intelligent systems.

Organizers: Senya Polikovsky


  • Kevin T. Kelly
  • Max Planck House Lecture Hall

In machine learning, the standard explanation of Ockham's razor is to minimize predictive risk. But prediction is interpreted passively---one may not rely on predictions to change the probability distribution used for training. That limitation may be overcome by studying alternatively manipulated systems in randomized experimental trials, but experiments on multivariate systems or on human subjects are often infeasible or immoral. Happily, the past three decades have witnessed the development of a range of statistical techniques for discovering causal relations from non-experimental data. One characteristic of such methods is a strong Ockham bias toward simpler causal theories---i.e., theories with fewer causal connections among the variables of interest. Our question is what Ockham's razor has to do with finding true (rather than merely plausible) causal theories from non-experimental data. The traditional story of minimizing predictive risk does not apply, because uniform consistency is often infeasible in non-experimental causal discovery: without strong and implausible assumptions, the probability of erroneous causal orientation may be arbitrarily high at any sample size. The standard justification for causal discovery methods is point-wise consistency, or convergence in probability to the true causes. But Ockham's razor is not necessary for point-wise convergence: a Bayesian with a strong prior bias toward a complex model would also be point-wise consistent. Either way, the crucial Ockham bias remains disconnected from learning performance. A method reverses its opinion in probability when it probably says A at some sample size and probably says B incompatible with A at a higher sample size. A method cycles in probability when it probably says A, then probably says B incompatible with A, and then probably says A again. Uniform consistency allows for no reversals or cycles in probability. Point-wise consistency allows for arbitrarily many. Lying plausibly between those two extremes is straightest possible convergence to the truth, which allows for only as many cycles and reversals in probability as are necessary to solve the learning problem at hand. We show that Ockham's razor is necessary for cycle-minimal convergence and that patience, or waiting for nature to choose among simplest theories, is necessary for reversal-minimal convergence. The idea yields very tight constraints on inductive statistical methods, both classical and Bayesian, with causal discovery methods as an important special case. It also provides a valid interpretation of significance and power when tests are used to fish inductively for models. The talk is self-contained for a general scientific audience. Novel concepts are illustrated amply with figures and simulations.

Organizers: Michel Besserve Kun Zhang


(Matter) Waves in disordered media

Talk
  • 16 July 2015 • 15:00 15:30
  • Valentin Volchkov
  • AGBS Seminar Room

The propagation of waves in inhomogeneous media is a vast subject, spanning many different research communities. The ability of waves to interfere leads to the celebrated phenomenon of Anderson localization. Constructive interference increases the probability of return and therefore it can reduce or even cancel the propagation in a disordered medium. Anderson localization was first predicted for electrons in 'dirty' condensed matter systems, very soon however, it was generalized to all kind of waves and has been studied since with light, microwaves, ultrasound, and ultra cold atoms. Here I will give a brief introduction into the basic ideas of Anderson physics and mention some applications. In fact, I will argue that disorder can be used as a resource rather being a nuisance. I will discuss ultra cold atoms as a good candidate for studying Anderson localization and wave propagation in disorder in general and present related experiments.

Organizers: Senya Polikovsky


  • Garrett Stanley
  • MRZ Seminar room

The external world is represented in the brain as spatiotemporal patterns of electrical activity. Sensory signals, such as light, sound, and touch, are transduced at the periphery and subsequently transformed by various stages of neural circuitry, resulting in increasingly abstract representations through the sensory pathways of the brain. It is these representations that ultimately give rise to sensory perception. Deciphering the messages conveyed in the representations is often referred to as “reading the neural code”. True understanding of the neural code requires knowledge of not only the representation of the external world at one particular stage of the neural pathway, but ultimately how sensory information is communicated from the periphery to successive downstream brain structures. Our laboratory has focused on various challenges posed by this problem, some of which I will discuss. In contrast, prosthetic devices designed to augment or replace sensory function rely on the principle of artificially activating neural circuits to induce a desired perception, which we might refer to as “writing the neural code”. This requires not only significant challenges in biomaterials and interfaces, but also in knowing precisely what to tell the brain to do. Our laboratory has begun some preliminary work in this direction that I will discuss. Taken together, an understanding of these complexities and others is critical for understanding how information about the outside world is acquired and communicated to downstream brain structures, in relating spatiotemporal patterns of neural activity to sensory perception, and for the development of engineered devices for replacing or augmenting sensory function lost to trauma or disease.

Organizers: Jonas Wulff


Autonomous Systems At Moog

Talk
  • 06 July 2015 • 14:00 15:00
  • Gonzalo Rey
  • AMD Seminar Room

The talk will briefly introduce Moog Inc. It will then describe Moog's view of its value proposition to robotics and autonomous systems. If robots and autonomous system are to achieve their enormous potential to positively impact the world economy, the technology has to achieve equivalent the levels of robustness, availability, reliability and safety that are expected from current solutions. The commercial aircraft industry has seen an order of magnitude increase in machine complexity in the last fifty years in order to reach the highest ever levels of cost per seat-mile and safety in its history. Today one can travel cheaper and safer than ever. Moog believes that there are opportunities to apply the methodologies and principles that enabled the lowest ever costs while at the same time managing the highest ever complexity and safety levels for aircraft to robotics and autonomous systems. The talk will briefly describe the type of approaches used in aircraft to achieve such low levels of failures that are hard to comprehend (or believe for those not familiar with the engineering approach), while at the same time, relying on low cost commercial off the shelf components in electronics, materials and manufacturing processes. Next the talk will move onto a couple of active research projects Moog is engaged in with ETHZ and IIT. Finally, it will give an overview of an emerging research effort in certification of advanced (robot) control laws.

Organizers: Ludovic Righetti


  • Trevor Darrell
  • MPH Lecture Hall, Tübingen

Learning of layered or "deep" representations has provided significant advances in computer vision in recent years, but has traditionally been limited to fully supervised settings with very large amounts of training data. New results show that such methods can also excel when learning in sparse/weakly labeled settings across modalities and domains. I'll present our recent long-term recurrent network model which can learn cross-modal translation and can provide open-domain video to text transcription. I'll also describe state-of-the-art models for fully convolutional pixel-dense segmentation from weakly labeled input, and finally will discuss new methods for adapting deep recognition models to new domains with few or no target labels for categories of interest.

Organizers: Jonas Wulff


  • Andre Seyfarth
  • MRZ Seminar Room

In this presentation a series of conceptual models for describing human and animal locomotion will be presented ranging from standing to walking and running. By subsequently increasing the complexity of the models we show that basic properties of the underlying spring-mass model can be inherited by the more detailed models. Model extensions include the consideration of a rigid trunk (instead of a point mass), non-elastic leg properties (instead of a mass-less leg spring), additional legs (two and four legs), leg masses, leg segments (e.g. a compliantly attached foot) and energy management protocols. Furthermore we propose a methodology to evaluate and refine conceptual models based on the test trilogy. This approach consists of a simulation test, a hardware test and a behavioral comparison of biological experiments with model predictions and hardware models.


  • Andre Seyfarth
  • MRZ Seminar room

In this presentation a series of conceptual models for describing human and animal locomotion will be presented ranging from standing to walking and running. By subsequently increasing the complexity of the models we show that basic properties of the underlying spring-mass model can be inherited by the more detailed models. Model extensions include the consideration of a rigid trunk (instead of a point mass), non-elastic leg properties (instead of a mass-less leg spring), additional legs (two and four legs), leg masses, leg segments (e.g. a compliantly attached foot) and energy management protocols. Furthermore we propose a methodology to evaluate and refine conceptual models based on the test trilogy. This approach consists of a simulation test, a hardware test and a behavioral comparison of biological experiments with model predictions and hardware models.

Organizers: Ludovic Righetti