Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Machine Learning ❤️ Google

Machine Learning ❤️ Google

I gave this talk on how to get started with Machine Learning with Google products like TensorFlow and the Vision API on DevFest Paraná (29/11), Belo Horizonte (05/11) and São Paulo (12/11).

Avatar for Maria Clara Santana

Maria Clara Santana

November 12, 2016
Tweet

More Decks by Maria Clara Santana

Other Decks in Programming

Transcript

  1. Como escrever um código para diferenciar uma de uma ?

    (aka hello world de machine learning)
  2. > tweet.json { "document":{ "type":"PLAIN_TEXT", "content":"Want to recover a locked

    AppleID and then Apple asks if I want to have an unlock email sent to the locked AppleID. Fuck logic." } }
  3. { "documentSentiment": { "polarity": -1, "magnitude": 1.4 }, "language": "en"

    } $ curl -s -X POST -H "Content-Type: application/json" --data- binary @tweet.json https://language.googleapis.com/v1beta1/ documents:analyzeSentiment?key=${NL_API_KEY}
  4. { "entities": [ { "name": "Apple", "type": "ORGANIZATION", "metadata": {

    "wikipedia_url": "https://en.wikipedia.org/wiki/ Apple_Inc." }, "salience": 0.18, }, ... ] } $ curl -s -X POST -H "Content-Type: application/json" --data- binary @tweet.json https://language.googleapis.com/v1beta1/ documents:analyzeEntities?key=${NL_API_KEY}
  5. > tweet.json { "document":{ "type":"PLAIN_TEXT", "content":"Want to recover a locked

    AppleID and then Apple asks if I want to have an unlock email sent to the locked AppleID. Fuck logic." }, "features":{ "extractSyntax":true } }
  6. $ curl -s -X POST -H "Content-Type: application/json" --data- binary

    @request.json https://language.googleapis.com/v1beta1/ documents:annotateText?key=${NL_API_KEY}
  7. > audio.json { "config": { "encoding":"FLAC", "sample_rate": 16000, "language_code": "en-US"

    }, "audio": { "uri":"gs://cloud-samples-tests/speech/brooklyn.flac" } } Áudio: http://bit.ly/2eSp4Z6
  8. $ curl -s -X POST -H "Content-Type: application/json" --data- binary

    @audio.json "https://speech.googleapis.com/v1beta1/ speech:syncrecognize?key=${SPEECH_API_KEY}" { "results": [ { "alternatives": [ { "transcript": "how old is the Brooklyn Bridge", "confidence": 0.98267895 } ] } ] }
  9. Transfer Learning / Inductive Learning + Inception V3 API Leitura

    adicional: https://arxiv.org/abs/1512.00567
  10. % cd $HOME % mkdir tf_files % cd tf_files %

    curl -O http://download.tensorflow.org/example_images/ flower_photos.tgz % tar xzf flower_photos.tgz % docker run -it -v $HOME/tf_files:/tf_files gcr.io/tensorflow/ tensorflow:latest-devel # cd /tensorflow # git pull
  11. # python tensorflow/examples/image_retraining/retrain.py \ --bottleneck_dir=/tf_files/bottlenecks \ --how_many_training_steps 500 \ --model_dir=/tf_files/inception

    \ --output_graph=/tf_files/retrained_graph.pb \ --output_labels=/tf_files/retrained_labels.txt \ --image_dir /tf_files/flowers_photos
  12. % cd $HOME/tf_files % curl -L https://goo.gl/tx3dqg > $HOME/tf_files/label_image.py %

    docker run -it -v $HOME/tf_files:/tf_files gcr.io/ tensorflow/tensorflow:latest-devel% mkdir paintings % python /tf_files/label_image.py /tf_files/flower_photos/ daisy/21652746_cc379e0eea_m.jpg daisy (score = 0.99071) sunflowers (score = 0.00595) dandelion (score = 0.00252) roses (score = 0.00049) tulips (score = 0.00032)