Research

DeepSeek for Research? Abstract Recently released weight open model deepseek-v3 and deepseek-r1 have brought a new wave of AI fever. Impressive user experience of deepseek-r1 has once again sparked the AI craze. In this report, we will introduce the characteristics of deepseek-r1, the details of the principles behind it, and its innovative points and differences from existing LLM models. After that, this report will further explore how deepseek-r1 can help academic research, introduce the role of prompt engineering in deepseek-r1, existing tools and solutions, and the problems faced by LLM in assisting academic research. ...

写在前面 MTCNN主页： https://kpzhang93.github.io/MTCNN_face_detection_alignment/index.html MTCNN论文： https://arxiv.org/abs/1604.02878 MTCNN，采用级联CNN结构，抓住了人脸检测和对齐这两个任务之间内在的相关性。同时做到输出人脸的Bounding Box以及人脸的Landmark(眼睛、鼻子、嘴)的位置。 ...