Jan Sochman

WaldBoost - learning for time constrained sequential detection

In many computer vision classification problems, both the error and time characterizes the quality of a decision. We show that such problems can be formalized in the framework of sequential decision-making. If the false positive and false negative error rates are given, the optimal strategy in terms of the shortest average time to decision (number of measurements used) is the Walds sequential probability ratio test (SPRT). We built on the optimal SPRT test and enlarge its capabilities to problems with dependent measurements. We show how to overcome the requirements of SPRT - (i) a priori ordered measurements and (ii) known joint probability density functions. We propose an algorithm with near optimal time and error rate trade-off, called WaldBoost, which integrates the AdaBoost algorithm for measurement selection and ordering and the joint probability density estimation with the optimal SPRT decision strategy. The WaldBoost algorithm is tested on the face detection problem. The results are superior to the state-of-the-art methods in the average evaluation time and comparable in detection rates.

A system for real-time detection and tracking of vehicles from a single car-mounted camera

A novel system for detection and tracking of vehicles from a single car-mounted camera is presented. The core of the system are high-performance vision algorithms: the WaldBoost detector [1] and the TLD tracker [2] that are scheduled so that a real-time performance is achieved. The vehicle monitoring system is evaluated on a new dataset collected on Italian motorways which is provided with approximate ground truth (GT0) obtained from laser scans. For a wide range of distances, the recall and precision of detection for cars are excellent. Statistics for trucks are also reported. The dataset with the ground truth is made public.

Who knows who - Inverting the Social Force Model for finding groups

Social groups based on friendship or family relations are very common phenomena in human crowds and a valuable cue for a crowd activity recognition system. In this paper we present an algorithm for automatic on-line inference of social groups from observed trajectories of individual people. The method is based on the Social Force Model (SFM) - widely used in crowd simulation applications - which specifies several attractive and repulsive forces influencing each individual relative to the other pedestrians and their environment. The main contribution of the paper is an algorithm for inference of the social groups (parameters of the SFM) based on analysis of the observed trajectories through attractive or repulsive forces which could lead to such behaviour. The proposed SFM-based method shows its clear advantage especially in more crowded scenarios where other state-of-the-art methods fail. The applicability of the algorithm is illustrated on an abandoned bag scenario.

AdaBoost with totally corrective updates for fast face detection

An extension of the AdaBoost learning algorithm is proposed and brought to bear on the face detection problem. In each weak classifier selection cycle, the novel totally corrective algorithm reduces aggressively the upper bound on the training error by correcting coefficients of all weak classifiers. The correction steps are proven to lower the upper bound on the error without increasing computational complexity of the resulting detector. We show experimentally that for the face detection problem, where large training sets are available, the technique does not overfit. A cascaded face detector of the Viola-Jones type is built using AdaBoost with the totally corrective update. The same detection and false positive rates are achieved with a detector that is 20% faster and consists of only a quarter of the weak classifiers needed for a classifier trained by standard AdaBoost. The latter property facilitates hardware implementation, the former opens scope for the increease in the search space, e.g the range of scales at which faces are sought.

Robust abandoned object detection integrating wide area visual surveillance and social context

This paper presents a video surveillance framework that robustly and efficiently detects abandoned objects in surveillance scenes. The framework is based on a novel threat assessment algorithm which combines the concept of ownership with automatic understanding of social relations in order to infer abandonment of objects. Implementation is achieved through development of a logic-based inference engine based on Prolog. Threat detection performance is conducted by testing against a range of datasets describing realistic situations and demonstrates a reduction in the number of false alarms generated. The proposed system represents the approach employed in the EU SUBITO project (Surveillance of Unattended Baggage and the Identification and Tracking of the Owner).

Learning a fast emulator of a binary decision process

Computation time is an important performance characteristic of computer vision algorithms. This paper shows how existing (slow) binary-valued decision algorithms can be approximated by a trained WaldBoost classifier, which minimises the decision time while guaranteeing predefined approximation precision. The core idea is to take an existing algorithm as a black box performing some useful binary decision task and to train the WaldBoost classifier as its emulator. Two interest point detectors, Hessian-Laplace and Kadir-Brady saliency detector, are emulated to demonstrate the approach. The experiments show similar repeatability and matching score of the original and emulated algorithms while achieving a 70-fold speed-up for Kadir-Brady detector.

Inter-stage feature propagation in cascade building with AdaBoost

A modification of the cascaded detector with the Ada-Boost trained stage classifiers is proposed and brought to bear on the face detection problem. The cascaded detector is a sequential classifier with the ability of early rejection of easy samples. Each decision in the sequence is made by a separately trained classifier, a stage classifier. In proposed modification the features from one stage of training are propagated to the next stage classifier. The proposed intra-stage feature propagation is shown to be greedily optimal, does not increase computational complexity of the stage classifier and leads to shorter stage classifiers and accordingly to faster detectors. A cascaded face detector is built with the intra-stage feature propagation and is compared with the Viola and Jones approach. The same detection and false positive rates are achieved with a detector that is 25% faster and consists of only two thirds of the weak classifiers needed for a cascade trained by the Viola and Jones approach. The latter property facilitates hardware implementation, the former opens scope for the increase in the search space, e.g., the range of scales at which faces are sought.

Learning Fast Emulators of Binary Decision Processes

Computation time is an important performance characteristic of computer vision algorithms. The paper shows how existing (slow) binary decision algorithms can be approximated by a (fast) trained WaldBoost classifier.WaldBoost learning minimises the decision time of the classifier while guaranteeing predefined precision. We show that the WaldBoost algorithm together with bootstrapping is able to efficiently handle an effectively unlimited number of training examples provided by the implementation of the approximated algorithm.Two interest point detectors, the Hessian-Laplace and the Kadir-Brady saliency detectors, are emulated to demonstrate the approach. Experiments show that while the repeatability and matching scores are similar for the original and emulated algorithms, a 9-fold speed-up for the Hessian-Laplace detector and a 142-fold speed-up for the Kadir-Brady detector is achieved. For the Hessian-Laplace detector, the achieved speed is similar to SURF, a popular and very fast handcrafted modification of Hessian-Laplace; the WaldBoost emulator approximates the output of the Hessian-Laplace detector more precisely.

Training sequential on-line boosting classifier for visual tracking

On-line boosting allows to adapt a trained classifier to changing environmental conditions or to use sequentially available training data. Yet, two important problems in the on-line boosting training remain unsolved: (i) classifier evaluation speed optimization and, (ii) automatic classifier complexity estimation. In this paper we show how the on-line boosting can be combined with Waldpsilas sequential decision theory to solve both of the problems. The properties of the proposed on-line WaldBoost algorithm are demonstrated on a visual tracking problem. The complexity of the classifier is changing dynamically depending on the difficulty of the problem. On average, a speedup of a factor of 5-10 is achieved compared to the non-sequential on-line boosting.

Wald's Sequential Analysis for Time-constrained Vision Problems

In detection and matching problems in computer vision, both classification errors and time to decision characterize the quality of an algorithmic solution. We show how to formalize such problems in the framework of sequential decisionmaking and derive quasi-optimal time-constrained solutions for three vision problems. The methodology is applied to face and interest point detection and to the RANSAC robust estimator. Error rates of the face detector proposed algorithm are comparable to the state-of-the-art methods. In the interest point application, the output of the Hessian-Laplace detector [9] is approximated by a sequential WaldBoost classifier which is about five times faster than the original with comparable repeatability. A sequential strategy based on Wald’s SPRT for evaluation of model quality in RANSAC leads to significant speed-up in geometric matching problems.