note.wcoder.com
wcoder GitHub

Table of Contents

[TOC]

去雾算法

北大&北航提出FFA-Net去雾新网络和何凯明博士提出的暗通道去雾算法,现所有源码已开源。
其论文链接:https://arxiv.org/abs/1911.07559。

https://github.com/zhilin007/FFA-Net

Python 还能实现图片去雾?FFA 去雾算法、暗通道去雾算法用起来!

拼图

https://github.com/shivaverma/Jigsaw-Solver
使用神经网络解决拼图游戏

图片操控

论文地址:https://arxiv.org/pdf/2007.06600.pdf

代码地址:https://github.com/genforce/sefa

项目主页:https://genforce.github.io/sefa/

AI图片生成器

https://github.com/AUTOMATIC1111/stable-diffusion-webui

https://github.com/InstantID/InstantID

demo: https://huggingface.co/spaces/InstantX/InstantID

图像分类

https://github.com/floydhub/imagenet

目标检测

https://github.com/amdegroot/ssd.pytorch

Image Style Transfer

https://github.com/zhanghang1989/PyTorch-Multi-Style-Transfer

Image-to-Image Translation

https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix

ImageCaptioning描述

https://github.com/ruotianluo/ImageCaptioning.pytorch

情感分析

https://github.com/bentrevett/pytorch-sentiment-analysis

问答

https://github.com/allenai/document-qa

动作跟随

https://github.com/facebookresearch/AnimatedDrawings

https://github.com/aigc3d/motionshop
https://aigc3d.github.io/motionshop/

翻译

https://github.com/OpenNMT/OpenNMT-py

文本转语音

Podcastfy

能够智能地将网页、PDF、图片等多种格式的内容转换成吸引人的多语言播客对话,是Google NotebookLM播客功能的开源替代品。

https://github.com/souzatharsis/podcastfy

换脸

https://github.com/AliaksandrSiarohin/first-order-model
视频教程:https://openbayes.com/console/openbayes/public/containers/BwZQj5wr3Jp

https://github.com/deepfakes/faceswap

DeepNude 脱衣

在.\deepnude_official-master\checkpoints 下载三个数据集cm.lib、mm.lib、mn.lib
下载地址:http://39.105.149.229/dn.7z

input.png的要求:
必须是*.png
必须是512px*512px

“蚂蚁呀嘿” https://github.com/alievk/avatarify

人脸识别

Face_Recognition API
EagleEye
DocFace
Get Me Through
DeepCamera

AI绘画

Midjourney(简称MJ),Stable-Diffusion(简称SD)
MJ需要付费使用,而SD开源免费(上手难度和学习成本略大,并且非常吃电脑配置(显卡、内存))

2020技术突破

自然语言处理(NLP)

最大语言模型GPT-3

论文地址:
https://arxiv.org/abs/2005.14165
项目地址:
https://github.com/openai/gpt-3
参考链接:
https://openai.com/blog/openai-api/

最大的称号现在是MT-NLG

GPT-J

https://github.com/kingoflolz/mesh-transformer-jax

EleutherAI Open-Sources Six Billion Parameter GPT-3 Clone GPT-J

最大聊天机器人BlenderBot

官方博客:

https://ai.facebook.com/blog/state-of-the-art-open-source-chatbot/

项目地址:
https://parl.ai/projects/recipes/

计算机视觉(CV)

纯视觉自动驾驶算法Sparse4D

https://github.com/linxuewu/Sparse4D

目标检测模型DETR

官方博客:
https://ai.facebook.com/research/publications/end-to-end-object-detection-with-transformers

源代码:
https://github.com/facebookresearch/detr

语义分割模型FasterSEG

论文地址:
https://arxiv.org/abs/1912.10917
项目地址:
https://github.com/VITA-Group/FasterSeg

EfficientDet-D7

论文地址:
https://arxiv.org/abs/1911.09070

项目地址:
https://github.com/google/automl/tree/master/efficientdet

Detectron2

开源地址:
https://github.com/facebookresearch/detectron2

DeepMind的AlphaFold解决蛋白质折叠问题

开源代码:
https://github.com/deepmind/deepmind-research/tree/master/alphafold_casp13
论文地址:
https://www.biorxiv.org/content/10.1101/846279v1.full.pdf

强化学习

Agent57得分高于人类baseline

论文地址:
https://arxiv.org/abs/1207.4708f

机器学习运维兴起

MLOps(Machine learining Operations)
参考链接:
https://nealanalytics.com/expertise/mlops/
开源地址:
https://github.com/microsoft/MLOps

AI对抗新冠病毒

新冠病毒九大研究方向:

https://www.who.int/blueprint/priority-diseases/key-action/Global_Research_Forum_FINAL_VERSION_for_web_14_feb_2020.pdf?ua=1
新冠数据集:
https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge/

知识库、知识图谱

事实 + 规则 = 知识库

逻辑编程

KnowledgeEditor

https://github.com/nicola-decao/KnowledgeEditor
真正的知识库不仅仅是‘问答’
EMNLP'21| KnowledgeEditor:编辑语言模型中的事实知识

← Previous Next →
Less
More