R-CNN author Ross Girshick resigned, He Kaiming and Xie Saining returned to academia, how many great minds have emerged from Meta CV

Yann LeCun said: "It is our loss that talents leave FAIR, but I am still happy for them."

Another big-name research scientist has left, this time Ross Girshick, the author of R-CNN

Recently, Meta chief scientist Yann LeCun announced on Twitter, Ross Girshick will leave FAIR to join the Allen Institute for Artificial Intelligence (AI2). Previous resignations include ResNeXt author Xie Saining (joined New York University as an assistant professor), Georgia Gkioxari (joined Caltech as an assistant professor), etc.

^{Source: https://twitter.com/ylecun/status/1730713022195470541}

us I checked Ross Girshick's personal homepage and confirmed his resignation from FAIR. He will join AI2 in early 2024.

Ani Kembhavi, senior director of computer vision at AI2, said Ross Girshick will join the PRIOR team. PRIOR stands for Perceptual Reasoning and Interaction Research. It is the computer vision research team of AI2 and is committed to advancing computer vision research to create AI systems that can see, explore, learn and reason about the world

^{Source: https://twitter.com/anikembhavi/status/1730655170038821085}

##Ross Girshick posted an article recalling his career at Meta, saying FAIR was and will remain an amazing place. But staying in one place for too long (8 years) may be a good reason to leave. Reinitialization and randomization are very important in a research career. In addition, he also stated that any talk about publishing indicators is pure nonsense.

Please refer to the following link for image source: https://twitter.com/inkynumbers/status/1730735493711810639

In fact, in addition to the fact that

He Kaiming announced his return to academia at the end of July this year and will join the EECS Department of Electrical Engineering and Computer Science at the Massachusetts Institute of Technology (MIT) in 2024, FAIR has been Many big names in the CV field have emerged.

Yann LeCun said that their departure is a loss to FAIR, but he is happy for them. He sees absolutely nothing wrong with scientists in industrial labs moving to academia or nonprofit organizations. For some, this is a natural career move.

Rewritten content: LeCun also used Bell Labs as an example. Many scientists at the lab will obtain tenure at a good university after leaving 5 to 10 years (skip completely) the difficult process of securing tenure). At different stages of life, priorities change. People who have been in industry for a long time may want to go into teaching, be with students, and enjoy the direct rewards of teaching

In fact, one can get into academia after working at FAIR for a few years of tenure, this is a feature, not a bug. This shift is possible at FAIR, which, like Bell Labs, practices open research and encourages scientists to publish.

The start of FAIR means that people can choose their career without taking any risks. This is a good thing for both practitioners and academics as it expands the research ecosystem

LeCun also pointed out that in recent years, many talented young computer scientists have chosen to join FAIR, such as Ishan Misra, Nicolas Carion The outflow and inflow of

talents such as , Xinlei Chen and Christoph Feichtenhofer is a normal thing, and many people choose to leave their comfort zone. However, some people believe that the continuous departure of AI giants from FAIR can shed some light on the current situation of the organization

Rewritten content is as follows: Image source: https://twitter.com/LearnOpenCV/status/1730736970136158274

In the past year, Meta has successively open sourced large-scale models of the Llama and Llama 2 series, becoming an indispensable force in the open source community. However, Meta also faces many challenges in retaining artificial intelligence talents, and the loss of talent is inevitable. Scientists like Ross Girshick who have extensive experience in industry moving to universities or non-profit institutions will bring a unique perspective to academia and potentially produce more impactful research

RBG Master: Introduction to Ross Girshick

Personal homepage link: https://www.rossgirshick.info/

Previously, Ross Girshick was a research scientist at Meta FAIR, working on computer vision and machine learning from 2015 to 2023. He received his PhD in computer science from the University of Chicago in 2012.

Before joining FAIR, Ross served as a researcher at Microsoft Research and a postdoctoral fellow at the University of California, Berkeley. There, he studied under Professors Jitendra Malik and Trevor Darrell

Ross’s research interests include visual perception algorithms (object recognition, localization, segmentation, pose estimation, etc.), representation learning (using pre-trained networks with strong supervision, weak supervision, or no supervision at all) and vision and language research.

For his contributions to open source software and datasets, Ross was awarded the 2017 PAMI Young Investigator Award, as well as the 2017, 2021, and 2023 PAMI Mark Everingham Awards

Ross Achieving many achievements in the field of artificial intelligence, he first became famous for developing the region-based convolutional neural network (R-CNN) object detection method. This research can be said to have completely changed the research direction in the field of target detection. Subsequent research such as Fast-RCNN and Faster-RCNN are all developed on the basis of R-CNN

His Google Scholar citation is currently Already more than 410,000 times

In the work Ross has participated in in the past, there are many popular research projects, such as Fast R-CNN, Mask R-CNN, YOLO, Faster R-CNN and SAM, etc.

In 2017, Mask R-CNN, which Ross participated in, won the ICCV Marr Award (Best Paper). This paper now has more than 30,000 citations; another paper "Focal Loss for Dense Object Detection" won the ICCV Best Student Paper that year.

In 2021, the paper "Masked Autoencoders Are Scalable Vision Learners" in which Girshick participated became a hot topic in the computer vision circle. This paper demonstrates a new method called masked autoencoders (MAE) that can be used as a scalable self-supervised learner for computer vision.

This year, Meta released the "Segment Anything" (Segment Anything) model (SAM), which has been hailed by many as a subversion of research in the traditional CV field. Ross is one of the authors of this paper. one.

Now I choose to go to AI2, and I look forward to more amazing works from Girshick.

The above is the detailed content of R-CNN author Ross Girshick resigned, He Kaiming and Xie Saining returned to academia, how many great minds have emerged from Meta CV. For more information, please follow other related articles on the PHP Chinese website!