Qifeng Chen

Assistant Professor at HKUST

[Linus ✦ Ekenstam @LinusEkenstam]

CoDeF 🤯

Temporal Consistent Video Processing 🫨

This is game over, I can not believe what I just saw.

This is ControlNet for a whole video 😱

Links in next tweet

[AK @_akhaliq]

Blind Video Deflickering by Neural Filtering with a Flawed Atlas

project page: https://t.co/BStkbNmOZ7

[Qifeng Chen @CQFHK]

7 papers accepted to #ICCV2021! Thanks to our hard-working HKUST students and collaborators. The works are mostly about new discoveries in computational photography and autonomous driving.

[Qifeng Chen @CQFHK]

@AutoVisionGroup @Mi_Niemeyer Congratulations!

[Qifeng Chen @CQFHK]

😊x7 #CVPR2021

[Georgia Gkioxari @georgiagkioxari]

The list of paper IDs provisionally accepted to #CVPR2021 is available online. Congratulations to all accepted papers!

https://t.co/nsPJCnrjxF

The list will appear tomorrow on the #CVPR2021 website. Bye-bye DAF's server!

[Qifeng Chen @CQFHK]

Our AI Application for Music Videos: Music to Dance Video Generation https://t.co/tRbvMVQlKx

[Qifeng Chen @CQFHK]

We present a new triple-camera system that can obtain high-resolution depth maps beyond 300 meters https://t.co/CuT8SszrwN

[Reza Zadeh @Reza_Zadeh]

Learning to See in the Dark: fast imaging in low light. Operates directly on raw sensor data, replacing the traditional image processing pipeline with a CNN. Does better than increasing exposure & ISO, or a flash.

Paper: https://t.co/1J87JDWqgj
Code: https://t.co/pDHOZr7iME

[Qifeng Chen @CQFHK]

We propose a new semi-parametric image synthesis approach that combines the strengths of deep learning and nonparametric external memory
https://t.co/3hsmLeEv3F

[Qifeng Chen @CQFHK]

We use a context aggregation network with adaptive normalization to accelerate advanced image processing operators. https://t.co/x7C5rEbeDw

[Qifeng Chen @CQFHK]

@alexbrattmd The training loss I use is a perceptual loss which is designed for natural images. We may need to design a different loss for medical scans.

[Qifeng Chen @CQFHK]

@HellDunkelkeit @SashaVNovikov @karpathy These two semantic layouts are not part of the training set.

[Qifeng Chen @CQFHK]

@alexbrattmd I am solving the inverse problem of semantic segmentation. This problem is not trivial. There is no ground truth given a semantic layout.

[Qifeng Chen @CQFHK]

We synthesize photographic images at 2-megapixel resolution. No GANs. https://t.co/vBPsqL2gwD

[Qifeng Chen @CQFHK]

New to twitter!