Skip to content

S0soo/tf-faster-rcnn-ghostnet

Repository files navigation

tf-faster-rcnn-ghostnet

This is a demo about Faster R-CNN use GhostNet

Explain

This is a demo of Faster R-CNN for Windows. The backbone network uses GhostNet.

Faster R-CNN code from dBeker

GhostNet code from huawei-noah

This Demo is only for learning and communication, it is not certain that the code is completely correct. Due to limited computing resources, no suitable hyperparameters were found, and results did not perform well.

Dependence

python 3.6
tensorflow 1.8.0
tensorpack 0.9.7

How To Use

You need to make a dataset in VOC format

If it doesn't work, you need to set it up here and replace it with the corresponding file in the library

 1. Put the dataset into './data'

 2. in './lib/datasets/pascal_voc.py' to change class

 <self.path = r'D:\Faster-R-CNN\labels_13_BJOYYX.txt'>

 3. Set pre-training file path

 <pretrained_model = r'F:\GhostNet\ghostnet\models\ghostnet_checkpoint'>

run train-ghostnet.py to train

run test-ghostnet.py to test

Results

In my own dataset, batch=128, iteration=100000, results:

result

The results are not ideal, and I will continue to search for more suitable hyperparameters.
GhostNet is tiny.

About

This is a demo about Faster R-CNN use GhostNet

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published