Videos with results of our activities

The videos present miscellaneous results of lab's research, projects, and other activities. Most videos present lab's publications, while some show final results of our projects.

Video restoration and teeth segmentation

Video Thumbnail
Click to play

Our MOST-Net method is a multi-task deep neural architecture. Without frame latency, it processes noisy and blurry video frames to produce enhanced videos with object segmentation masks.

Efficient video
denoising

Video Thumbnail
Click to play

Our BP-EVD algorithm denoises VGA videos in real-time and without frame latency. The results show DAVIS sequences with Gaussian noise and after denoising.

Joint video denoising and deblurring

Video Thumbnail
Click to play

Our R2-D4 algorithm jointly denoises and deblurs videos of regular scenes. The demo show results for static and moving camera with increasing levels of noise.

Dual PTZ camera
cooperation

Video Thumbnail
Click to play

The dual camera system robustly tracks a moving target. A wide-field camera retargets the narrow-field camera when the zoomed object leaves its view. The visual tracker is continuously supported by an object detector.

The videos present miscellaneous results of lab's research, projects, and other activities. Most videos present lab's publications, while some show final results of our projects.

The videos present miscellaneous results of lab's research, projects, and other activities. Most videos present lab's publications, while some show final results of our projects.

Video restoration and teeth segmentation

Odtwórz wideo

Our MOST-Net method is a multi-task deep neural architecture. Without frame latency, it processes noisy and blurry video frames to produce enhanced videos with object segmentation masks.

Efficient video
denosising

Odtwórz wideo

Our BP-EVD algorithm denoises VGA videos in real-time and without frame latency. The results show DAVIS sequences with Gaussian noise and after denoising.

Joint video denoising and deblurring

Odtwórz wideo

Our R2-D4 algorithm jointly denoises and deblurs videos of regular scenes. The demo show results for static and moving camera with increasing levels of noise.

Dual PTZ camera
cooperation

Odtwórz wideo

The dual camera system robustly tracks a moving target. A wide-field camera retargets the narrow-field camera when the zoomed object leaves its view. The visual tracker is continuously supported by an object detector.