MEFLUT: Unsupervised Learning 1D Lookup Tables for Multi-exposure Image Fusion
Ting Jiang, Chuan Wang, Xinpeng Li, Ru Li, Haoqiang Fan, Shuaicheng Liu

International Conference on Computer Vision （ICCV） 2023 | Paper | Code

• This work introduce MEFLUT that learns 1D LUTs for the task of MEF. It show that the fusion weights can be encoded into the LUTs successfully. Once learned, MEFLUT can be easily deployed with high efficiency, so that a 4K image runs in less than 4ms on a PC GPU.
• This is the first time to demonstrate the benefits of LUTs for MEF.

SAM-IQA: Can Segment Anything Boost Image Quality Assessment?
Xinpeng Li, Ting Jiang, Haoqiang Fan, Shuaicheng Liu

ArXiv 2023 | Paper| Code

• This work first introduce the SAM encoder as a feature extractor in IQA and demonstrate its strong generalization ability in this domain

Realistic Noise Synthesis with Diffusion Models
Qi Wu, Mingyan Han, Ting Jiang, Haoqiang Fan, Bing Zeng, Shuaicheng Liu

ArXiv 2023 | Paper

• This work first propose a real noise data synthesis approach based on the diffusion model.
• A camera setting embedding approach is designed that can better control the distribution and level of generated noise.

DIPNet: Efficiency Distillation and Iterative Pruning for Image Super-Resolution
Lei Yu, Xinpeng Li, Youwei Li, Ting Jiang, Qi Wu, Haoqiang Fan, Shuaicheng Liu

IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW) 2023 | Paper | Code

• Winner method of the NTIRE Super-Resolution Challenge 2023.
• This work propose the use of enhanced HR images to improve the learning ability of lightweight networks.
• A novel multi-stage lightweight training strategy combining distillation, progressive learning, and pruning is proposed.

Efficient Neural Architecture Search via firefly optimization for SAR ship detection
Jielei Wang, Zongyong Cui, Ting Jiang, Changjie Cao, Zongjie Cao

IEEE Transactions on Image Processing (TIP) 2022 | Paper

• A set of lightweight detection networks for SAR ship target detection are proposed.

ADNet: Attention-guided Deformable Convolutional Network for High Dynamic Range Imaging
Zhen Liu, Wenjie Lin, Xinpeng Li, Ting Jiang, Mingyan Han, Qing Rao, Haoqiang Fan, Jian Sun, Shuaicheng Liu

IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW) 2021 | Paper | Code | [Video: Youtube, Bilibili]

• This work proposes a novel dual-branch pipeline for multiframe HDR imaging of dynamic scenes.

Filter pruning with a feature map entropy importance criterion for convolution neural networks compressing
Jielei Wang, Ting Jiang, Zongyong Cui, Zongjie Cao

Neuro Computing 2021 | Paper

• A novel structured pruning method for Convolutional Neural Networks (CNN) compression is proposed, where filter-level redundant weights are pruned according to entropy importance criteria (termed FPEI).

Driver Monitor System

• Dec. 2018 - Jul. 2019, Major Membor
• Responsible for designing a light-weighted end-to-end network running on an onboard chip to detect smoking, phoning, and drinking in driving.
• I participated in the overall pipeline of algorithm development, including data collection, data cleaning, algorithm development, etc.

High Dynamic Range Image (HDRI)

• Jul. 2019 - Present, Algorithm Leader
• Responsible for HDRI projects, including YUV and RAW domains. Compared with ordinary images, HDRI can provide more dynamic range and image details, which can better reflect the visual effects in the real environment of people.

Super Night View

• Jul. 2019 - Present, Algorithm Leader
• Responsible for Super night view projects, including YUV and RAW domains. In hand-held super night scene, improve night scene brightness, clarity, noise, dynamic range and other performance.

Multi-frame or Single-frame Denoising

• Oct. 2021 - Present, Algorithm Leader
• Responsible for image and video denoising projects, including YUV and RAW domains. In extreme scenarios, the noise is improved, and it is suitable for various mobile devices, such as mobile phones, vehicles, drones, robots, etc.

Super Resolution (SR)

• Jul. 2022 - Present, Algorithm Leader
• Responsible for image and video SR projects, including YUV and RAW domains. The recovery effect of details in the smart phone super resolution project is better than that of the competitor.

Portrait Beauty

• Jul. 2019 - Nov. 2019, Algorithm Leader
• Responsible for leading Portrait beauty projects. As a team leader, I led team members in developing a novel beautifying algorithm capable of adaptive face area adjustment to achieve effective portrait enhancement. This algorithm was successfully applied to products and widely adopted.

Object Detection

• Jul. 2021 - Mar. 2022, Algorithm Leader
• Responsible for Object Detection projects. It mainly involves the detection of ghosts existing in the project to achieve precise positioning.

Intelligent Erase

• Sep. 2021 - Mar. 2022, Algorithm Leader
• Responsible for Intelligent Erase projects. The recovery effect of details in the smart erasing project is better than that of competing models.