Reduce the model complexity by 612 times, and memory footprint by 19.5 times compared to base model, while achieving worst case accuracy threshold.
lightweight
deep-learning
tensorflow
keras
cnn
mnist
inception
ensemble-model
model-complexity
model-compression
memory-footprint
mobile-platform
model-acceleration
1x1
conv1x1
-
Updated
Jan 26, 2018 - Jupyter Notebook