resource.md

Resources

Table of Contents

• Resources
  • PYTHON
    • LIBRARIES
  • TENSORFLOW
    • GITHUB
    • INSTALL
    • TUTORIALS
      • KERAS
      • BAZEL
  • Google Cloud Platform
  • CONTENT
  • EXAMPLES
  • Computer Vision and Pattern Recognition
    • Current papers
    • News

---

PYTHON

• Versions download
  • https://www.python.org/downloads/
  • https://www.python.org/downloads/mac-osx/

LIBRARIES

• Data Visualization, Graphing Library, https://matplotlib.org/
• Numbers, Arrays, Calc, Formulas, https://numpy.org/
• https://www.tensorflow.org/guide/keras

TENSORFLOW

• https://www.tensorflow.org/

GITHUB

• https://github.com/tensorflow/tensorflow

INSTALL

• https://www.tensorflow.org/install/pip

TUTORIALS

• https://www.tensorflow.org/tutorials
• [Audio recognition using spectrograms]](https://www.tensorflow.org/tutorials/sequences/audio_recognition)

KERAS

• https://keras.io/#getting-started-30-seconds-to-keras

BAZEL

• https://bazel.build
• https://docs.bazel.build/versions/master/install-os-x.html

Google Cloud Platform

• https://cloud.google.com/
  • Upper right corner, sign in using your Google acct.
  • Upper right, select console
  • Upper left Navigation Menu

CONTENT

• Ornithology image library

EXAMPLES

• Fashion MNIST Discern between shirts, shoes

Computer Vision and Pattern Recognition

Applying model on Mobile device

Once model is built, can be tested on mobile device.

• Android Camera Demo

Xcode and Swift Topics

• (Image Classification)[https://developer.apple.com/documentation/vision/classifying_images_with_vision_and_core_ml]

Current papers

• https://arxiv.org/list/cs.CV/recent

News

• Facial Recog in 'chimps'
• AI System to Locate Vehicles per geography and time from traffic cams, given sample image

---

WORKFLOWS

1. TF output
2. .pb file/package
3. statistical graph representation of the data set

TENSORFLOW TO BUILD MODEL

• Demo load images using tf.data and a URL

MANAGING RAW IMAGE SET

For training and testing:

1. Images all same size, same orientation horiz or vert
2. Put copyright in images or metadata
3. Select 80%-90% images for training model, remaining to test against

• some with some without animals

• https://www.tensorflow.org/tutorials/images/classification

• http://download.tensorflow.org/example_images/flower_photos.tgz

*https://docs.docker.com/storage/bind-mounts/

• snap docker py 2.7 needs pip2.7 install matplotlib