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

Hoi Lam - Adding ML Kit to Android Things And s...

Hoi Lam - Adding ML Kit to Android Things And some TensorFlow things

Avatar for droidcon Berlin

droidcon Berlin

July 17, 2018
Tweet

More Decks by droidcon Berlin

Other Decks in Programming

Transcript

  1. #WeAreHoi #WeAreHoi Adding ML Kit to Android Things And some

    TensorFlow things DroidCon Berlin 2018
  2. #WeAreHoi Image Labelling Low latency High accuracy 400+ labels 10,000+

    labels Free Free then quota Check the website!
  3. #WeAreHoi • Control your environment (apples and oranges) • Make

    your model smaller - Quantization - instead of using Float32 use uint8 ◦ Need fake quantization node ◦ Search for Fixed Point Quantization for more details • Mixing samples ◦ java.lang.IllegalArgumentException: Failed to get input dimensions. 0-th input should have 602112 bytes, but found 150528 bytes. ◦ java.lang.IllegalArgumentException: Cannot convert an TensorFlowLite tensor with type FLOAT32 to a Java object of type [[B (which is compatible with the TensorFlowLite type UINT8)
  4. #WeAreHoi #WeAreHoi Adding ML Kit to Android Things And some

    TensorFlow things DroidCon Berlin 2018