├── pose.md └── README.md /pose.md: -------------------------------------------------------------------------------- 1 | # Pose estimation 2 | 3 | This page provides an overview on how to get started quickly with deep learning models for pose estimation on your own computer. 4 | 5 | ## Python3 and Tensorflow 6 | I would propose that we try to work in the same environment as much as possible, to facilitate sharing of knowledge and work. 7 | There are several deep-learning frameworks, of which Tensorflow is currently probably the most widespread. 8 | In addition, *Keras* is a relatively easy to use higher-level library that works on top of tensorflow. 9 | 10 | Installing Tensorflow with *GPU* support is not strictly necessary, but will greatly improve learning and inference speed. The installation 11 | is a little more involved than the CPU alternative, so you may want to start with that. You do need a graphic card that supports CUDA for 12 | this, like most nvidia cards. To my knowledge, GPU support is not available on macOS. (But correct me if I'm wrong) (JanM: there are ways...) 13 | 14 | ### Linux 15 | ...todo... 16 | 17 | ### Mac 18 | The procedure described on the [Tensorflow macOS install page](https://www.tensorflow.org/install/install_mac) is pretty straightforward, the virtualenv method works fine. 19 | 20 | GPU support is difficult, see [this guide](https://byai.io/howto-tensorflow-1-6-on-mac-with-gpu-acceleration/). Also it is rather pointless because Tensorflow requires CUDA which has no support on factory MacBooks, because no NVidea GPUs. 21 | 22 | ### Windows 23 | For Windows, Anaconda is a convenient way to install Python, Tensorflow and many other Python libraries. As stated on the 24 | [Tensorflow website](https://www.tensorflow.org/install/install_windows) though, the Anaconda package of Tensorflow is community supported, 25 | not officially supported. But in my experience pip install does not allways work well on Windows when packages need to be compiled. So 26 | Anaconda provides an easy to use installer. 27 | 28 | To get started, download the python3.6 distribution from [https://www.anaconda.com/download/](https://www.anaconda.com/download/). 29 | Then follow the guide on [https://www.tensorflow.org/install/install_windows](https://www.tensorflow.org/install/install_windows) 30 | 31 | ## Pose estimation models 32 | 33 | #### tf-pose-estimation 34 | A nice pretrained model to begin with is [tf-pose-estimation](https://github.com/handsonai/tf-pose-estimation). 35 | Clone or download the repository, then run 36 | ``` 37 | $ python src/run_webcam.py --model=mobilenet_thin --resolution=432x368 --camera=0 38 | ``` 39 | Note that the version linked here is adapted from the original, a run_vectors.py script was added by Fako Berkers that outputs json with the coordinates of the pose. 40 | 41 | ##### macOS install 42 | Depending on what you currently have installed, the install in the readme of tf-pose-estimation might not work rightaway. 43 | I had to edit requirements.txt and remove the lines for ast and dill (it still seems to work afterwards). 44 | openCV is needed, I already had it installed so I don't know what is required exactly. 45 | ``` 46 | brew install opencv 47 | ``` 48 | To get the webcam demo running (live video input), I had to install ffmpeg with x265 support, in my case that meant: 49 | ``` 50 | brew reinstall ffmpeg --with-x265 51 | ``` 52 | To get the image-file demo running (run.py), I got error from matplotlib about Python needing to be installed as a Framework, but [there is always StackOverflow](https://stackoverflow.com/a/21789908/403991). 53 | 54 | 55 | #### Openpose 56 | To quickly get started on Windows you can download the [binaries](https://github.com/CMU-Perceptual-Computing-Lab/openpose/releases) for the [openpose project](https://github.com/CMU-Perceptual-Computing-Lab/openpose). These allow you to use their openpose system with a webcam without having to install anything else (they come with a .bat file to download the needed models). 57 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # resources-overview 2 | Overview of resources on AI and related topics. 3 | 4 | # Format next meetup 5 | - Show & Tell 6 | - World Café (table discussions) 7 | 8 | Check the Slack: https://handsonai.slack.com/, ask Johan or Jan for an invite. 9 | 10 | # Mentioned in Slack 11 | Machine Learning for Artists Github: https://github.com/ml4a 12 | 13 | Generating Kittens: http://affinelayer.com/pixsrv/index.html 14 | 15 | # Topics from the meetup 16 | hacking events (on specific techniques), improvised music, text, neural networks, jobs, media, image recognition, image generation, generative art, emotions in text, emotions in face images, micro-expressions, design, warable, installation art, film, video, photography. 17 | 18 | What can I do with AI? 19 | 20 | Prototype things and see how they impact behavior. 21 | 22 | AI answering big life questions: should I have children? 23 | 24 | AI writing interactive horror story (whatsapp bots). 25 | 26 | AI penpall (fiction). 27 | 28 | List open source AI libs and services. 29 | 30 | All AI use same "theory of emotion" <-> philosophy of algorithms. 31 | 32 | Training data: what to use, where to get it, who owns it? 33 | 34 | AI as a new perspective on things/problems/solutions (not as the solution itself). 35 | 36 | Neural feedback loop: AI recognizing emotions and sending feedback to make user aware of self (mindfullness). 37 | 38 | Create images from words (like "Platonic", or rather Wittgensteinian, examples). 39 | 40 | "Hipsterbar" digital doorkeeper (and ethical implications). 41 | 42 | Human computing and gamification (metadata and trainingset creation, for example for Europeana): duolingo, recaptcha, RNA sequencing games, galaxy zoo, crowd crafting, mechanical turk, crowdflower. 43 | 44 | Types of AI: supervised learning, unsupervised learning, rule based, natural computing, game AI, ... 45 | 46 | Ethics, futurism, http://lesswrong.com/ 47 | 48 | Algorithms 101 49 | 50 | Kaggle competition hack evenings: https://www.kaggle.com/ 51 | 52 | # Possible presentations 53 | - Introduction to Natural Computing (by Johan Groenen) 54 | - Introduction to Artificial Intelligence (by Johan Groenen) 55 | 56 | # Online courses 57 | https://www.udacity.com/course/intro-to-machine-learning--ud120 58 | 59 | https://www.coursera.org/learn/machine-learning 60 | 61 | https://www.udacity.com/course/intro-to-statistics--st101 62 | 63 | https://www.udacity.com/course/deep-learning--ud730 64 | 65 | # Youtube links 66 | Deep Learning: https://www.youtube.com/channel/UC9OeZkIwhzfv-_Cb7fCikLQ 67 | 68 | Neural Network: https://www.youtube.com/watch?v=AyzOUbkUf3M 69 | 70 | # Websites 71 | V2_ website: http://v2.nl/events/hands-on-ai-kick-off 72 | 73 | Meetup page: https://www.meetup.com/Unstable-Media-Meetup-Rotterdam/ 74 | 75 | # Alex Falks notes 76 | 77 | ## Creative applications 78 | - Horror story bot 79 | - Improv music 80 | - AI "jamming" 81 | - Chat/text robots 82 | - AI playing games 83 | - Improving "single player mode" 84 | 85 | ## Ethical implications 86 | - Ethics 87 | - Privacy & Transparency 88 | - Data collection 89 | - AI Rebels 90 | - We don't know what we don't know 91 | 92 | ## AI categories 93 | - Unsupervised machine learning 94 | - Supervised machine learning 95 | - Euristics 96 | - Swarm Algoritm 97 | - Semantic Toolboxes 98 | - Expert systems 99 | - Rule based systems 100 | - Robotics 101 | 102 | ## Social applications 103 | - AI psychology 104 | - Medical data analysis 105 | - Neural feedback 106 | - Handicap applications (AI for blind people) 107 | - City oriented bots 108 | 109 | ## Human/Computer Integration 110 | - AI/Human Synergy 111 | - Thought provoking bots 112 | - Brain/computer Interface 113 | - Cyborgs 114 | 115 | ## AI Community 116 | - AI History 117 | - Discourse group 118 | - Collectively owned AI resource 119 | - Talent development 120 | - Open source collection 121 | - Shared data sets 122 | - Skill exchange 123 | 124 | ## Imagery 125 | - Pattern recognition 126 | - Image generation 127 | - recognizing chairs 128 | 129 | ## Education 130 | - Where to look/start? 131 | - Low threshold 132 | - Making machine learning fun! 133 | --------------------------------------------------------------------------------