tf-mobilenet-v2

Mobilenet V2(Inverted Residual) Implementation & Trained Weights Using Tensorflow

Training

$ python3 run.py train --num_gpu=1 --depth_multiplier=1.0 --datadir=... 

Pretrained Models

I trained in a few ways, but I failed to replicate the result from the original paper. (2~4% Accuracy Drop)

But you can use the pretrained weight in tensorflow now : https://github.com/tensorflow/models/tree/master/research/slim#pre-trained-models

References

• Inverted Residuals and Linear Bottlenecks: Mobile Networks for Classification, Detection and Segmentation
• Paper Review (Korean)
• Mobilenet V2 Tensorflow Implementation : https://github.com/timctho/mobilenet-v2-tensorflow
• Mobilenet V2 Pytorch Implementation : https://github.com/MG2033/MobileNet-V2
• Mobilenet V1 Official Implementation
• Mobilenet V1 Official Test
• Checkmate : Tensorflow Drop-in Saver