Skip to content

Latest commit

 

History

History
24 lines (17 loc) · 971 Bytes

File metadata and controls

24 lines (17 loc) · 971 Bytes

LightFlow Tensorflow/Keras implementation

Implementation of Optical Flow predictions using deep learning LightFlow architecture proposed in Towards High Performance Video Object Detection for Mobiles

The main training procedure is based on https://github.com/sampepose/flownet2-tf , and the network was builded using Keras.

Tensorbaord logs

lightflow tensorboard Loss lightflow tensorboard Val image

Actual input/output

inputs inputs

lightflow Actual output

TODOs

  • Fix error with the BatchNorm layer in Keras, model does not converge.
  • Migrate Keras model definition to TensorFlow
  • Work in progress

Reference article, thanks