wadecong/Scene_Recognition

Scean_Recognition for Place365

Installation

if you want to use pretrained models, then all you need to do is:

git clone https://github.com/djang000/Scene_Recognition.git

if you also want to train new modes, you will need natural images for training and MobileNet wegihts by running.

you can download MobileNet_V2 weight from below site

https://github.com/tensorflow/models/tree/master/research/slim/nets/mobilenet

To prepare the Place365 dataset for use with train_mobilenet_V2.py, you can download Place365 dataset I used place365-standard dataset for training (105GB with 256 x 256 images).

if you download Place365 dataset, put train image and meta directory to data folder.

http://places2.csail.mit.edu/download.html

Usage

Following are examples of how the scripts in this repo can be used.

• Scene_eval.ipynb

  you can show the evaluation result using trained model.

• train.py

  you should run below command for training.

  python train_mobilenet_V2.py

Env

conda create -n tensorflow_v1 python=3.7
conda activate tensorflow_v1
pip install "tensorflow==1.15"
conda install jupyter notebook
conda install scipy
conda install matplotlib