Task 1: Multi-scale Object Detection

Task 2: Long-term Multi-object Tracking

Task 3: Group Classification

798,000,000

Pixels (Max)

13

Natural Life Scenes

104,196

Head points

121,027

Bounding Boxes

News

GigaCamera System Introduction

UnstructuredCam

Gigapixel videography, beyond the resolution of single camera and human visual perception, plays an important role in capturing large-scale dynamic scene with extremely high resolution for both macro and micro domains. Restricted by the spatial-temporal bandwidth product of optical system, the size, weight, power and cost are central challenges in gigapixel video.

The UnstructuredCam we designed, an end-to-end unstructured multi-scale camera system, shows the ability of real-time capture, dynamically adjusting local-view cameras, and online warping for synthesizing gigapixel video. We take the advantage of our UnstructuredCam to develop the Tsinghua Gigapixel Dataset. And these datasets we provide in www.gigacamera.com are all characterized by extremely high resolution, large scale and wide FOV. We hope that our datasets will help researchers explore the cutting-edge computer vision tasks such as long-term multi-target object tracking, large-scale crowd counting, large-scale generic object detection, etc. » Continue Reading

Fig.1 Illustration of representative imaging systems. (a) single camera imaging system faces the contradiction between wide FOV and high resolution, (b) single-scale camera array imaging [1][2] relies on image stitching[3], (c) structured multi-scale camera array (AWARE2[4]) adopts two-stage optical imaging design, (d) un-structured multi-scale camera array (denoted as UnstructuredCam).

Fig.2 Schematic and point spread function (PSF) of RUSH.

RUSH Macroscopy

Large-scale imaging of biological dynamics at high spatiotemporal resolution is indispensable in the study of system biology. However, with conventional microscopes, one has to make a compromise between large field-of-view (FOV) and high spatial resolution, resulting from the inherently limited space-bandwidth-product (SBP). In addition, no imaging system is of sufficient data throughputs to record such huge information yet. Here we break these bottlenecks by proposing the flat-curved-flat strategy, in which the sample plane is magnified into a large spherical image surface and then is seamlessly conjugated to multiple planar sensors with a relay lens array. Accordingly, we develop a customized objective of globally-uniform 0.92 μm resolution across a 10 mm×12 mm FOV, and an accompanying camera array for high-throughput recording at 5.1 giga-pixels per second. We demonstrate the first reported video-rate, giga-pixel imaging of biological dynamics at centimeter scale and micron resolution, including brain-wide structural imaging and functional imaging in awake, behaving mice. » Continue Reading

Project Demo Videoes

Live Demo of UnstructuredCam

Live Demo of RUSH Macroscopy:
Brain-wide functional imaging in awake mice with music stimuli.

Dataset Image Gallery

• Sequence ID: OCT Bay
• Image size: 26753 x 15052
• Frame rate: 30 Hz

• Sequence ID: Shanghai Train Station
• Image size: 26558 x 14828
• Frame rate: 30 Hz

• Sequence ID: Nanshan I Park
• Image size: 32609 x 24457
• Frame rate: 12 Hz

• Sequence ID: Primary School
• Image size: 31760 x 23810
• Frame rate: 12 Hz

• Sequence ID: Basketball Court
• Image size: 31753 x 23810
• Frame rate: 12 Hz

• Sequence ID: HIT Canteen
• Image size: 26753 x 15052
• Frame rate: 30 Hz

• Sequence ID: Shanghai Marathon
• Image size: 26908 x 15024
• Frame rate: 30 Hz

• Sequence ID: Shenzhenbei Station
• Image size: 26583 x 14957
• Frame rate: 30 Hz

• Sequence ID: Shenzhen Library
• Image size: 32129 x 24096
• Frame rate: 30 Hz

• Sequence ID: Xili Pedestrian Street
• Image size: 26583 x 14957
• Frame rate: 30 Hz

• Sequence ID: Xili Crossroad
• Image size: 26753 x 15052
• Frame rate: 30 Hz

Professors in Our Team

Lu Fang

Tsinghua University

Qionghai Dai

Tsinghua University

David J. Brady

Duke University
Top