Shifeng Zhang
PhD Candidate
Haidian District, Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolShifeng Zhang is a fifth-year bachelor-to-PhD candidate at the National Laboratory of Pattern Recognition, CASIA, specializing in machine learning and pattern recognition with an emphasis on object, face, pedestrian and video detection. Guided by Prof. Stan Z. Li, he has eight years of hands-on experience translating research into robust detection systems and real-time models. His open-source contributions include maintenance and testing of widely cited detectors such as RefineDet, FaceBoxes and ATSS (CVPR/IJCB), where he implemented loss refinements, TOPK positive sample selection and practical integration fixes for VGG16 deployments. He combines experimental rigor with engineering pragmatism—improving dataset handling, test scripts and model configuration to boost reproducibility and deployment readiness. Based in Haidian, Beijing, he bridges academic research and production-ready code, often focusing on edge cases and evaluation pipelines that are easy to overlook.
8 years of coding experience
CASIA
Github Skills (16)
face-detection10
object-detection10
computer-vision10
pytorch10
machine-learning10
caffe10
python10
testing10
algorithms9
ml-deployment9
vggnet9
data-structure9
continuous-deployment9
algorithm9
data-structures9
Programming languages (2)
Github contributions (5)
sfzhang15/ATSS
Dec 2019 - Apr 2020
Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selection, CVPR, Oral, 2020
Role in this project:
ML Engineer
ML EngineerContributions:14 commits, 10 pushes, 98 comments in 4 months
Contributions summary:Shifeng primarily focused on modifications related to the loss function within the ATSS (Anchor-free Training Sample Selection) framework. Their commits demonstrate the implementation of a "TOPK" positive sample selection strategy and address bugs related to edge cases. These changes are targeted at improving the model's performance and robustness, demonstrating a strong understanding of object detection concepts.
pytorchdeep-learningadaptivecomputer-visionanchor-free
sfzhang15/RefineDet
Nov 2017 - Mar 2019
Single-Shot Refinement Neural Network for Object Detection, CVPR, 2018
Role in this project:ML Engineer
ML EngineerContributions:104 commits, 11 PRs, 79 pushes in 1 year 4 months
Contributions summary:Shifeng contributed to the refinement and testing of a Single-Shot Refinement Neural Network for object detection, specifically in the context of computer vision. The commits primarily focused on updating configuration files and model definitions related to RefineDet, including modifying the deployed models, and adapting testing and finetuning scripts for the VGG16 architecture. These changes included updates to pre-trained model paths and data loading configurations, demonstrating a focus on model integration and testing.
pytorchrefinementdeep-learningobject-detectioncomputer-vision
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.