计算机视觉

Search documents
ImageNet作者苏昊被曝任教复旦
量子位· 2025-10-10 03:52
Core Viewpoint - The article discusses the potential appointment of Hao Su, a prominent figure in embodied intelligence and computer vision, to Fudan University, highlighting his significant contributions to the field and his entrepreneurial ventures in robotics and simulation [1][49][51]. Group 1: Hao Su's Academic and Research Background - Hao Su is an associate professor at the University of California, San Diego (UCSD), specializing in computer vision, graphics, embodied intelligence, and robotics [14][49]. - He was involved in the creation of ImageNet and has led foundational projects such as ShapeNet, PointNet, and SAPIEN, which have significantly advanced the fields of 2D and 3D vision [4][30][34]. - Su's research has evolved from natural language processing to computer vision and then to 3D vision, culminating in the development of large-scale datasets and models that have transformed the landscape of artificial intelligence [22][30][34]. Group 2: Contributions to Robotics and Simulation - In 2020, Su launched SAPIEN, the first simulator focused on generalizable robotic operations, and later developed the ManiSkill platform for training robotic skills [35][41]. - His company, Hillbot, co-founded in 2024, aims to leverage high-fidelity simulation for robotics, with products like Hillbot Alpha designed for complex environments [43][45]. - Hillbot has partnered with Nvidia to generate high-quality training data, indicating a strong focus on enhancing robotic capabilities through advanced simulation techniques [47]. Group 3: Potential Move to Fudan University - There are rumors that Su will join Fudan University, which may invest in his company Hillbot and potentially appoint him to dual roles at various research institutes [51][52]. - Fudan University has established a credible embodied intelligence research institute, offering competitive salaries and performance-based incentives, which could attract top talent like Su [55][57].
算法小垃圾跳槽日记 2024&2025版
自动驾驶之心· 2025-10-06 04:05
以下文章来源于AIZOO ,作者元峰 AIZOO . 每日分享计算机视觉(CV)和深度学习的资讯、博文、教程,分享大佬的经验。CV圈的朋友们快快进来吧。 作者 | 元峰 来源 | AIZOO 原文链接: 分享一下近期换工作面试的一些事 点击下方 卡片 ,关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近30个 方向 学习 路线 >>自动驾驶前沿信息获取 → 自动驾驶之心知识星球 本文只做学术分享,如有侵权,联系删文 年前的时候换了个工作,从外卖厂离职了。离职之前,恰好有假期,基本上一个月都在疯狂的面试,平均一天都得六个面试,特此记录分享一下。 下图是 24年9月的面试日历,基本上整整一个月都在面试,我看了一下,有几天竟然是8场面试,真的是累的不轻,眼睛都很干涩。其实,面试的时候,精神的 高度聚焦的,这样一天下来,简直比上班都累。 | | 周一 | | 周二 | | 周三 | 周四 | 周五 | | 周六 | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | 三十 | 2日 八月初一 | | 3日 初二 | | 4日 初三 | ...
刚毕业的AI博士,滞销了
投资界· 2025-09-28 07:35
Core Viewpoint - The article highlights the stark contrast in the job market for AI PhDs, where top-tier talent is highly sought after and rewarded, while the majority of average AI PhDs struggle to find suitable employment opportunities, leading to a polarized job landscape [5][10]. Group 1: Job Market Dynamics - The job market for AI PhDs is characterized by a significant divide, with top-tier candidates receiving lucrative offers, while average candidates face rejection and limited opportunities [5][10]. - Companies are increasingly selective, with hiring ratios for desirable positions often exceeding 10:1, and in some cases, as high as 200:1 for certain roles [8][9]. - Many average AI PhDs find themselves in a "talent pool," waiting for opportunities that may never materialize, as they lack the necessary credentials or connections to secure positions [9][10]. Group 2: Recruitment Challenges - Average AI PhDs often struggle to meet the high expectations set by companies, which seek candidates with extensive publication records and relevant experience [12][21]. - The pressure to publish is immense, with some candidates feeling compelled to produce papers that may lack genuine innovation just to meet job application requirements [13][18]. - The recruitment process is lengthy and often results in disappointment, as candidates face long waiting periods only to receive rejection notices [7][8]. Group 3: The Role of Networking - Networking plays a crucial role in securing job opportunities, with many positions being filled through personal connections rather than solely based on qualifications [21][22]. - Companies often prefer candidates who come recommended by trusted sources, such as former professors or industry contacts, which can disadvantage those without such connections [21][22]. - The reliance on networking highlights the importance of building relationships within the industry, as many job openings are not publicly advertised [21][22]. Group 4: Industry Trends and Expectations - The AI industry is rapidly evolving, with a strong focus on commercial applications and the development of general-purpose models, which may not align with the specialized research backgrounds of many PhDs [18][20]. - Companies are increasingly looking for candidates who can contribute to immediate business needs rather than those with niche expertise, leading to a mismatch between academic training and industry requirements [19][20]. - The competitive landscape for AI talent is intensifying, with top companies offering attractive packages to lure the best candidates, further widening the gap for average PhDs [10][11].
美睫机器人:嫁接睫毛又快又好
Ke Ji Ri Bao· 2025-09-18 00:17
(文章来源:科技日报) 如今,很多爱美人士都会定期进行睫毛嫁接。以往人工嫁接睫毛需要两个多小时,而现在最新美睫机器 人能提供更高效的睫毛嫁接服务。 这款机器人集成了计算机视觉与人工智能等前沿技术,将原本长达两个多小时的睫毛嫁接过程缩短至20 分钟。美睫师先清洁客户睫毛,去除灰尘与油脂以保证嫁接持久度,接着给客户贴上带引导条码的眼贴 膜,为机器人定位提供参考。准备就绪后,美睫机器人开始操作。借助计算机视觉技术,它能够精准扫 描每位用户的眼部轮廓,还可以根据面部肌肉微动作进行实时动态调整,从而为用户定制独一无二的美 睫方案。同时,它采用的AI算法可将睫毛嫁接位置的误差控制在10微米以内。 ...
苹果首款智能眼镜聚焦无屏设计 预计12至16个月内推出
Huan Qiu Wang Zi Xun· 2025-09-15 04:20
事实上,关于苹果智能眼镜的发布时间,此前已有郭明錤(Ming-Chi Kuo)等分析师提出可能在2026年 或2027年推出,而古尔曼是这一时间线的坚定支持者之一。他曾多次报道,苹果智能眼镜大概率于2026 年发布,同时也不排除因技术优化等因素推迟至2027年的可能性。 来源:环球网 【环球网科技综合报道】9月15日消息,据Applelnsider报道,彭博分析师马克・古尔曼近日在一档播客 节目中对苹果智能眼镜研发进展展开分析,指出苹果计划在未来12至16个月内推出首款智能眼镜产品, 且该产品将采用无显示屏设计,定位与Meta的Ray-Bans(雷朋)智能眼镜展开竞争。 据介绍,这款无屏智能眼镜将配备摄像头与具备播放、录音功能的音频系统,不过需连接iPhone才能实 现数据处理。而消费者期待的、能通过镜片显示内容的完整智能眼镜体验,目前仍需数年时间才能落 地。古尔曼解释,设备微型化与减重技术是当前核心瓶颈,苹果的目标是让智能眼镜尽可能接近普通轻 量眼镜形态,避免重蹈Apple Vision Pro因体积和重量带来的使用体验局限。 在市场竞争层面,古尔曼特别强调苹果在智能眼镜领域拥有天然优势。一方面,依托苹果强 ...
复旦微电:FPGA系列产品的应用,尚未涉及向脑机接口领域开拓
Ge Long Hui· 2025-09-12 09:29
格隆汇9月12日丨复旦微电(688385.SH)在投资者互动平台表示,公司FPGA芯片可应用于通信、工业控 制及高可靠领域,并正在积极拓展计算机视觉、机器学习、高速数字处理等应用场景。目前,公司 FPGA系列产品的应用,尚未涉及向脑机接口领域开拓。 ...
Diebold Nixdorf (NYSE:DBD) 2025 Conference Transcript
2025-09-10 17:32
Diebold Nixdorf Conference Call Summary Company Overview - **Company**: Diebold Nixdorf (NYSE: DBD) - **Core Markets**: Banking and Retail [6][7] - **Opportunities**: Focus on branch automation in banking and expansion in the U.S. retail market [6][8] Banking Segment Insights - **Branch Efficiency**: Banks are seeking to improve branch efficiency as running a branch network accounts for 60% of their expenses [7][23] - **Solutions Offered**: - Recycling ATMs and teller cash recyclers to enhance cash management [7][24] - Software solutions that integrate physical branches with digital channels [25] - **Productivity Gains**: Recent implementations have improved teller productivity by 50% [24] Retail Segment Strategy - **Market Position**: Strong presence in Europe, aiming to expand in the U.S. [7][8] - **Self-Checkout Leadership**: Achieved number one position in self-checkout in Europe, targeting U.S. market expansion [26][39] - **AI-Driven Solutions**: Focus on reducing theft at checkout through AI and computer vision technologies [28][30] Financial Performance and Projections - **Free Cash Flow**: Positive free cash flow for three consecutive quarters, with a target of $800 million by 2027 [10][12] - **Revenue Growth**: Projected mid-single-digit growth from low single digits, with a target of 4% to 6% top-line growth by 2027 [12][42] - **Backlog**: Current backlog stands at $980 million, providing strong visibility into future revenue [46] Operational Improvements - **Margin Growth**: Product margins improved from low teens to mid to high 20s over two years [18][48] - **Lean Manufacturing**: Continuous improvement initiatives have led to significant enhancements in quality and delivery times [49][50] Capital Allocation and Shareholder Returns - **Debt Management**: Aiming for a 1.5x net debt leverage ratio, maintaining a strong balance sheet [13] - **Share Buyback Program**: Announced a $100 million buyback program, with $38 million already executed [13][14] Risk Factors and Market Conditions - **Tariff Exposure**: Estimated impact of $5 million to $10 million from tariffs, mitigated by local-to-local manufacturing strategies [52][53] - **ATM Market Dynamics**: The ATM market is stable with a slight growth trend, primarily driven by replacement rather than new installations [32] Key Differentiators - **Comprehensive Solutions**: Integration of hardware, software, and services to enhance banking and retail operations [25][31] - **Customer-Centric Approach**: Focus on understanding customer needs and providing tailored solutions [39][40] Conclusion Diebold Nixdorf is positioned for growth in both banking and retail sectors, leveraging technology and operational efficiencies to enhance profitability and shareholder value. The company is committed to executing its strategic initiatives while maintaining a strong focus on free cash flow generation and capital allocation.
刚刚,李飞飞主讲的斯坦福经典CV课「2025 CS231n」免费可看了
机器之心· 2025-09-04 09:33
Core Viewpoint - Stanford University's classic course "CS231n: Deep Learning for Computer Vision" is officially launched for Spring 2025, focusing on deep learning architectures and visual recognition tasks such as image classification, localization, and detection [1][2]. Course Overview - The course spans 10 weeks, teaching students how to implement and train neural networks while gaining insights into cutting-edge research in computer vision [3]. - At the end of the course, students will have the opportunity to train and apply neural networks with millions of parameters on real-world visual problems of their choice [4]. - Through multiple practical assignments and projects, students will acquire the necessary toolset for deep learning tasks and engineering techniques commonly used in training and fine-tuning deep neural networks [5]. Instructors - The course features four main instructors: - Fei-Fei Li: A renowned scholar and Stanford professor, known for creating the ImageNet project, which significantly advanced deep learning in computer vision [6]. - Ehsan Adeli: An assistant professor at Stanford, focusing on computer vision, computational neuroscience, and medical image analysis [6]. - Justin Johnson: An assistant professor at the University of Michigan, with research interests in computer vision and machine learning [6]. - Zane Durante: A third-year PhD student at Stanford, researching multimodal visual understanding and AI applications in healthcare [7]. Course Content - The curriculum includes topics such as: - Image classification using linear classifiers - Regularization and optimization techniques - Neural networks and backpropagation - Convolutional Neural Networks (CNNs) for image classification - Recurrent Neural Networks (RNNs) - Attention mechanisms and Transformers - Object recognition, image segmentation, and visualization - Video understanding - Large-scale distributed training - Self-supervised learning - Generative models - 3D vision - Visual and language integration - Human-centered AI [16]. Additional Resources - All 18 course videos are available for free on YouTube, with the first and last lectures delivered by Fei-Fei Li [12].
计划2026年商业化应用!马斯克:特斯拉未来约80%价值将来自于Optimus擎天柱机器人【附人形机器人行业发展趋势】
Qian Zhan Wang· 2025-09-02 11:00
Group 1 - Elon Musk believes that approximately 80% of Tesla's future value will come from the Optimus robot [2] - The mission of the Optimus robot is to liberate human labor by taking over tedious or dangerous jobs, with plans for commercialization by 2026 [2][3] - Market sentiment is mixed, with a prediction that the likelihood of Optimus being launched before 2027 is only 40% according to Kalshi [3] Group 2 - The humanoid robot industry integrates advanced technologies from mechanical engineering, electronics, computer science, and artificial intelligence [3] - The Chinese humanoid robot market is projected to reach approximately 2.76 billion yuan in 2024, with significant growth expected by 2027 [4] - Global humanoid robot shipments are expected to reach 38,000 units by 2030 according to Qianzhan Industry Research Institute [5] Group 3 - Major tech companies and startups are actively pursuing mass production of humanoid robots, despite challenges such as high R&D costs and market acceptance [7] - The development of humanoid robots is expected to bring new productivity and lifestyle changes to society as technology advances and market demand grows [7]
2025年中国AI工业质检行业发展历程、产业链、市场规模、重点企业及未来趋势研判:AI工业质检市场规模快速增长,3C电子为最大应用领域[图]
Chan Ye Xin Xi Wang· 2025-08-30 01:02
Core Viewpoint - The AI industrial quality inspection (QI) sector is rapidly growing in China, driven by the integration of AI technologies such as machine vision and deep learning, which significantly enhance inspection efficiency and accuracy. The market size is projected to grow from 0.9 billion yuan in 2017 to 45.4 billion yuan in 2024, with a compound annual growth rate (CAGR) of 75.09% [1][13]. Industry Overview - AI industrial QI refers to the automated detection and identification of product quality in industrial production processes using AI technologies [1][13]. - Traditional quality inspection methods have been inefficient and inconsistent, particularly in precision manufacturing sectors like 3C electronics and automotive manufacturing [1][13]. Market Growth - The market for AI industrial QI in China is expected to reach 64.9 billion yuan by 2025, indicating continuous expansion driven by advancements in multi-modal detection technologies and deeper industry applications [1][13]. - The AI industrial QI market has transitioned from pilot applications to widespread adoption in high-end manufacturing sectors such as consumer electronics, new energy batteries, and semiconductors [1][13]. Technical Advantages - AI industrial QI systems offer high efficiency, accuracy, consistency, iterability, and data analysis capabilities, significantly improving the quality control process [5][6]. - The shift from classical machine learning algorithms to deep learning detection algorithms has reduced reliance on human analysis, enhancing the accuracy of defect detection [7]. Industry Chain - The AI industrial QI industry chain includes upstream components like machine vision software and hardware, optical devices, and image sensors, which are crucial for implementing AI QI applications [7][8]. - Downstream applications primarily involve sectors such as 3C electronics, automotive, lithium batteries, and semiconductors [7][8]. Image Sensor Market - The image sensor industry in China has seen rapid growth, with production expected to increase from 1.073 billion units in 2017 to 5.206 billion units in 2024, reflecting a CAGR of 25.31% [9][10]. - The market size for image sensors is projected to grow from 29.634 billion yuan in 2017 to 94.898 billion yuan in 2024, with a CAGR of 18.09% [9][10]. Downstream Market Structure - The 3C electronics sector dominates the AI industrial QI demand, accounting for over 50% of the market share, driven by the rapid development and innovation in consumer electronics [10][11]. - The automotive manufacturing sector holds a stable demand for AI industrial QI, representing 18.6% of the market share due to stringent quality control requirements [10][11]. Competitive Landscape - The AI industrial QI market in China is competitive with a low concentration, where the top five companies hold 44.7% of the market share [14]. - Key players include Baidu Group, Innovation Qizhi, and Tencent Cloud, with respective market shares of 10.6%, 10.4%, and 10.2% [14]. Future Trends - The AI industrial QI sector is expected to accelerate towards full automation, with deep learning-based visual inspection systems gradually replacing traditional manual inspections [16]. - There will be a continuous expansion of application scenarios, moving from established sectors to advanced manufacturing fields such as new energy and biomedicine [17]. - The integration of multi-modal technologies will enhance detection capabilities, allowing for comprehensive quality monitoring in complex industrial environments [18][19].