Learning to Track at 100 FPS with Deep Regression Networks

Abstract

Abstract. Machine learning techniques are often used in computer vision due to their ability to leverage large amounts of training data to improve performance. Unfortunately, most generic object trackers are still trained from scratch online and do not benefit from the large number of videos that are readily available for offline training. We propose a method for offline training of neural networks that can track novel objects at test-time at 100 fps. Our tracker is significantly faster than previous methods that use neural networks for tracking, which are typically very slow to run and not practical for real-time applications. Our tracker uses a simple feed-forward network with no online training required. The tracker learns a generic relationship between object motion and appearance and can be used to track novel objects that do not appear in the training set. We test our network on a standard tracking benchmark to demonstrate our tracker's state-of-the-art performance. Further, our performance improves as we add more videos to our offline training set. To the best of our knowledge, our tracker 1 is the first neural-network tracker that learns to track generic objects at 100 fps.

Tracking by Natural Language Specification

“…The common approach is to specify a target by means of a bounding box around the object and to track this target as it moves throughout the video [38,33,20]. The paradigm has proven to be effective and considerable progress has been achieved [17,37,34,3,11]. Yet, the fundamental assumption of having a bounding box target specification available has never been challenged.…”

Target Tracking via Particle Filter and Convolutional Network

“…In order to describe the target better, we applied convolutional networks to learn robust representations for visual tracking without offline training using a large amount of auxiliary data, which is inspired by recent studies [11,18]. First, we use predefined convolutional filters to extract the high-order features.…”

Visual Tracking Utilizing Object Concept From Deep Learning Network

“…In (Hong et al, 2015a), sampled feature maps are classified by SVM to generate a saliency map. More recently, a number of studies use Recurrent Neural Networks (RNNs) for visual tracking (Bertinetto et al, 2016, Held et al, 2016, Chen and Tao, 2016. In (Held et al, 2016), Held et.…”

“…In the first image, the target is cropped as target template with some background texture. Using the sample generation idea from (Held et al, 2016), we randomly shift and scale the target in the second image as search image to simulate the motion of the target and camera simultaneously. In tracking, we generally cropped a search image in the new frame based on the target's previous location instead to track the target in the whole image.…”

“…where u = 0, for both scale and translation changes for scale changes, b = bs = 1/5 for translation change, b = bt = 1/15, Also, we enforce the scale change to be less than ±0.4 and the center of the translated target is still in the search image similar to the work in (Held et al, 2016). The ImageNet dataset is mainly used to teach localization network to find object boundary and the smooth motion as complementary data in the limited ALOV300+ dataset.…”