How we use Deep Learning to classify more than 180,000 photos a day on PhotoGrid

  • R2
  • Day 2, 10:40‑11:10
  • Chinese talk w. English slides
  • Graphics
  • Novice

Deep Learning is so popular and in demand these days, and is becoming the state-of-the-art solution of many topics, such as image recognition, game playing, auto driving, etc.

Besides, we are so lucky that many Deep Learning frameworks and pre-trained models are just free and easy to use, that we can stand on the shoulders of giants and get great results.

In this talk, we are going to introduce how we use TensorFlow and Python to solve image classification problems with more than 180,000 photos a day on PhotoGrid, and avoid inappropriate content online.

We are going to have a brief introduction of Deep Learning and Transfer Learning, then share some experiences about the model training and performance tuning, and how we benefited from Deep Learning in practical use.

Talk Detail

### Python Tools Tensorflow : https://github.com/tensorflow/tensorflow Open source tools : https://github.com/cubrady/tf_inception_transfer_learning.git https://github.com/cubrady/parseWebsiteImage https://github.com/cubrady/py_cmds ### About PhotoGrid PhotoGrid is one of the most popular mobile photo editing apps of the world with more than 100 million user downloads. It's easy to create wonderful effects an share the wonderful moments to PhotoGrid community.

Slides Link

Speaker Information

Brad Chang

仍在軟體界打滾,初探深度學習領域中的一個迷途小書僮。
碰過幾種語言,但仍深深著迷於 Python 的快速簡潔易用。