Floris De Smedt

On-board real-time tracking of pedestrians on a UAV

Recent technical advances in Unmanned Aerial Vehicles (UAV) made a realm of applications possible. In this paper we focus on the application of following a walking pedestrian in real-time, using optimised pedestrian detection and object tracking. For this we use an on-board embedded system, offering an optimal ratio of computational power and weight. We extend the commonly used ground plane estimation technique, used to reduce the search space, based on the sensor data off the UAV. The integration of the ground plane constraint obtains a significant speed-up over the already optimised Aggregate Channel Feature (ACF) detector. To compensate for the frames without detections, we use a particle tracker based on color information. We successfully validated our system on a flying UAV.

Pedestrian Detection at Warp Speed: Exceeding 500 Detections per Second

Object detection, and in particular pedestrian detection, is a challenging task, due to the wide variety of appearances. The application domain is extremely broad, ranging from e.g. surveillance to automotive safety systems. Many practical applications however often rely on stringent real-time processing speeds combined with high accuracy needs. These demands are contradictory, and usually a compromise needs to be made. In this paper we present a pedestrian detection framework which is extremely fast (500 detections per second) while still maintaining excellent accuracy results. We achieve these results by combining our fast pedestrian detection algorithm (implemented as a hybrid CPU and GPU combination) with the exploitation of scene constraints (using a warping window approach and temporal information), which yields state-of-the-art detection accuracy. We present profound evaluation results of our algorithm concerning both speed and accuracy on the challenging Caltech dataset. Furthermore we present evaluation results on a very specific application showing the full potential of this warping window approach: detection of pedestrians in a trucks blind spot zone.

Faster and more intelligent object detection by combining OpenCL and KR

In this paper we present a fast implementation of a robust object detector by using OpenCL. The use of fast object detection is of great use for a broad range of applications in multiple domains. OpenCL allows for scalability to more performant and different types of hardware, with minimal changes to the implementation. By using a GPU as execution device, we exploit the data parallelism opportunities of the algorithm. We also discuss the use of knowledge representation as a means to integrate expert knowledge into applications. This can be used both for faster processing by limiting the searching space, and for applications to work more autonomous by exploiting a higher level of intelligence.

How to reach top accuracy for a visual pedestrian warning system from a car

Due to the wide applicability of pedestrian detection in surveillance and safety, this research topic has received much attention in computer vision literature. However, the focus of this research mainly lies in detecting and locating pedestrians individually as accurate as possible. In recent years, a number of datasets are captured using a forward looking camera from a car, which imposes the application of warning the driver when pedestrians are in front of the car. For such applications, it is not required to detect each pedestrian independently, but to generate an alarm when necessary. In this paper we explore techniques to boost the accuracy of recent channel-based algorithms in this application: algorithmic refinements as well as the inclusion of an LWIR image channel. We use the KAIST dataset which is constructed from image-pairs of both the visual and the LWIR spectrum, in day and night conditions. We study the influence of techniques that have shown success in literature.

Open Framework for Combined Pedestrian Detection

Pedestrian detection is a topic in computer vision of great interest for many applications. Due to that, a large amount of pedestrian detection techniques are presented in current literature, each one improving previous techniques. The improvement in accuracy in recent pedestrian detection, is commonly in combination with a higher computational requirement. Although, recently a technique was proposed to combine multiple detection algorithms to improve accuracy instead. Since the evaluation speed of this combination is dependent on the detection algorithm it uses, we provide an open framework that includes multiple pedestrian detection algorithms, and the technique to combine them. We show that our open implementation is superior on speed, accuracy and peak memory-use when compared to other publicly available implementations.

The Combinator: Optimal Combination of Multiple Pedestrian Detectors

In recent years, the accuracy of pedestrian detectors significantly improved. Currently, state-of-the-art pedestrian detectors achieve high accuracy results on challenging datasets. As opposed to refining a single detector, in this paper we propose a different approach to further increase the detection accuracy: combining multiple pedestrian detectors. The most straight-forward way to combine pedestrian detectors would be a naive AND or OR combination. Here, we present a novel generic combination framework in which we exploit specific information from each pedestrian detector to determine the optimal combination parameters. Our main motivation for this approach is based on the fact that several pedestrian detection approaches are based on very different techniques (e.g. a different feature pool), and thus an efficient combination should yield higher accuracy results. Indeed, such a combination is far more powerful, and our experiments indicate that specific (that is, cleverly chosen) combinations outperform existing state-of-the-art pedestrian detection results.

GPU Accelerated ACF Detector.

The field of pedestrian detection has come a long way in recent decades. In terms of accuracy, the current state-of-the-art is hands down reached by Deep Learning methods. However in terms of running speed this is not always the case, traditional methods are often still faster than their Deep Learning counterparts. This is especially true on embedded hardware, embedded platforms are often used in applications that require realtime performance while at same the time having to make do with a limited amount of resources. In this paper we present a GPU implementation of the ACF pedestrian detector and compare it to current Deep Learning approaches (YOLO) on both a desktop GPU as well as the Jetson TX2 embedded GPU platform.

Fast Rotation Invariant Object Detection with Gradient based Detection Models

Accurate object detection has been studied thoroughly over the years. Although these techniques have become very precise, they lack the capability to cope with a rotated appearance of the object. In this paper we tackle this problem in a two step approach. First we train a specific model for each orientation we want to cover. Next to that we propose the use of a rotation map that contains the predicted orientation information at a specific location based on the dominant orientation. This helps us to reduce the number of models that will be evaluated at each location. Based on 3 datasets, we obtain a high speed-up while still maintaining accurate rotated object detection.