轻舟已过万重山

原创 缺陷检测文献

缺陷检测文献阅读

原创 LLM笔记

原创 From Words to Worth: Newborn Article Impact Prediction with LLM笔记

类型（会议，期刊）图表数量。

原创 drawbacks of Diffusion Maps

both Diffusion Maps and many other manifold learning methods based on the spectraldecomposition of kernel matrices have drawbacks such as the inability to apply them to data outsidethe initial set

原创 Sensation, Perception and Cognition

Sensation: The brain’s raw response to a stimulus (e.g., feeling a vibration on your skin, seeing a flashing light).Perception: The brain’s interpretation of that sensation (e.g., recognizing that the flashing light is the letter “A”).Cognition: Higher-lev

原创 coherent motion

Author: ChatGPTCoherent motions refer to patterns of movement that are consistent, structured, and often correlated across space and/or time. The term appears across computer vision, robotics, and human motion analysis, with slightly different but related

原创 world models and Human–Object Interaction (HOI)

and。

原创 Human-Object Interaction (HOI) synthesis

Human-Object Interaction (HOI) synthesis refers to the generation of realistic interactions between humans and objects in virtual or physical environments. Unlike HOI recognition (which focuses on detecting existing interactions), HOI synthesis aims to cre

原创 “显著性”（Saliency）是计算机视觉中的一个重要概念，主要指的是图像或视频中最吸引人注意力的区域或对象

显著性”（Saliency）是计算机视觉中的一个重要概念，主要指的是。它模拟的是人类视觉系统对视觉场景中“显著”区域的感知能力。显著性可以用于图像理解、目标检测、图像压缩、图像分割等多个任务。

原创 Articulated objects are multi-part objects with joints that allow structured movement between parts.

that are。

原创 “Long-horizon” refers to interactions or sequences that unfold over an extended period of time

refers to。

原创 Papers about Anomaly Detection (Reconstruction-based and Restoration-based)

ICCV 2023 Institute of Automation, Chinese Academy of SciencesTargeting for detecting anomalies of various sizes fo rcomplicated normal patterns, we propose a Template-guided Hierarchical Feature Restoration method, which in-troduces two key techniques, b

原创 KNN Matting

香港科技大学CVPR 2012。

原创 Interactive Image Segmentation

CVPR 2023清华。

原创 Online Deep Clustering for Unsupervised Representation Learning

CVPR2020香港中文大学。

原创 Papers about Human in the Loop Annotation

ICCV 2021KAIST。

原创 Emergent Correspondence from Image Diffusion

Cornell UniversityFinding correspondences between images is a fundamental problem in computer vision. In this paper, we show that correspondence emerges in image diffusion models without any explicit supervision. We propose a simple strategy to extract thi

原创 weakly supervised semantic segmentation

CVPR2022 北京大学。

原创 Anomaly Detection Datasets

anomaly detection dataset

原创 GAN papers

gan papers

原创 papers about Semi-supervised

papers about semi-supervised

原创 Swin Transformer

Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. 2021AbtractThis paper presents a new vision Transformer, called Swin Transformer, that cap

原创 ViT预训练

CVPR 2022 何凯明。

原创 小样本分割 高斯过程

Johnander, J., Edstedt, J., Felsberg, M., Khan, F.S., Danelljan, M. (2022). Dense Gaussian Processes for Few-Shot Segmentation. In: Avidan, S., Brostow, G., Cissé, M., Farinella, G.M., Hassner, T. (eds) Computer Vision – ECCV 2022. ECCV 2022.Few-shot segme

原创 Convolution Network

https://www.youtube.com/watch?v=ANyxBVxmdZ0

原创 deep Bayesian learning

https://www.topbots.com/comprehensive-introduction-to-bayesian-deep-learning/NeurIPS 2020The key distinguishing property of a Bayesian approach is marginalization, rather than using a single setting of weights. Bayesian marginalization can particularly im

转载 Ranking loss

相关链接：https://gombru.github.io/2019/04/03/ranking_loss/https://github.com/adambielski/siamese-triplethttps://omoindrot.github.io/triplet-lossPyTorchCosineEmbeddingLoss. It’s a Pairwise Ranking Loss that uses cosine distance as the distance metric. Inp

原创 视觉SLAM十四讲__学习笔记

书里的代码: https://github.com/gaoxiang12/slambook。

原创 语义分割文献阅读笔记

语义 分割

原创 压缩感知文献阅读

针对正交匹配追踪(OMP)算法在压缩感知理论下的重构效果和所需时间相互矛盾的问题,基于子空间追踪(SP)算法的回溯思想,使用共轭梯度下降算法代替最小二乘法对正交匹配追踪(OMP)算法进行改进.并且对所改进算法的重构精度、重构稳定性进行了仿真实验,结果表明所提算法能保证重构质量良好并且有更好的重构速度和稳定性.马博珩,彭艺. 基于OMP算法的快速压缩感知图像重构[J]. 云南大学学报（自然科学版）,2017,39(2):207-211. DOI:10.7540/j.ynu.20160247.

原创 数学之美摘抄笔记

数学之美笔记

原创 Transformer文献阅读

文献阅读

原创 目标检测, 实例分割, 图像分类, panoptic segmentation文献

目标检测文献阅读

原创 Anomaly Detection | Metaformer

Paper： Learning Unsupervised Metaformer for Anomaly Detection

原创 anomaly detection dataset/papers/code

https://paperswithcode.com/dataset/mvtecadhttps://paperswithcode.com/dataset/kolektorsdd2

原创 CV顶会&顶刊

顶会：CVPR,ICCV,ECCV,ICML,NIPS, AAAI, IJCAI,BMVC顶刊：TPAMI, TIP, IJCV, JMLR

原创 The Correct Way to Measure Inference Time of Deep Neural Networks

https://deci.ai/resources/blog/measure-inference-time-deep-neural-networks/

转载 3. 确定性推理笔记

推理的基本概念3.1.1 推理的定义 3.1.2 推理方式及其分类 1.演绎推理：一般 → 个体 三段论式（三段论法） 2.归纳推理：个体 → 一般 完全归纳推理（必然性推理）不完全归纳推理（非必然性推理） 3.默认推理（缺省推理）：知识不完全的情况下假设某些条件已经具备所进行的推理。 1.确定性推理：推理时所用的知识与证据都是确定的，推出的结论也是确定的，其真值或者为真或者为假。 2.不确定性推理：推理时所用的知识与证据不都是确定的，推出的结论也是不确定的。 1.单调推理（基于经典

原创 人民日报训练word2vec实验

数据集人民日报：2020年10月04日-2021年10月04日概况25590 articles742362 sentences0.021 billion words294730 tokens182004942 pairs (window size: 5)词云(120 words)训练参数vector dimension: 100window size: 5K: 5batch size: 50epoch: 10learning rate: 0.025训练结果

原创 【文献阅读】Hybrid model for Chinese character recognition based on Tesseract-OCR

Tesseract-OCR engine+KNN+LSTMIntroductionChinese OCR is more difficultThe number of English letters is only 26. But the number of Chinese characters that used commonly are about 2,500.the strokes of Chinese characters are complex and similar.The dif