- Deep Learning based food recognition system
Using
Keras
(Tensorflow
backend) deep learning framwork - Also can be transplanted to other platforms like
Raspberry Pi
- This tensorflow repo is no longer maintained now :(, plz see FRS-pytorch for more details
- 1.run spiders
spider_baidu.py
orspider_douguo.py
to crawl raw image data from the internet - 2.create an empty folder and move raw images into it,in this project was
dataset
folder - 3.run
train.py
to train the model (only when dataset was downloaded)
- run
cam_demo.py
to show ui,load the model and recongnize the food
cd your project path and type:
python detect.py -i test.jpg
python detect.py -v test.mp4
- need plotting model structure? just install
graphviz
first - please screen out unqualified raw images manually after crawling
- file:
preprocess.py
- preprocess image dataset
- file:
image_util.py
- some image utils and algorithms
- file:
train.py
- main training program
- file:
cam_demo.py
,detect.py
- user interface,just to predict image,using pyqt5
- folder: spiders
- file:
spider_baidu.py
,spider_douguo.py
- use spiders to crawl raw images from the internet
$ pip install -r requirements.txt
- keras
- tensorflow-gpu
- numpy
- opencv-python
- pillow
- matplotlib (used to show parameter change)
- pyqt5 or wxpython (UI)
- graphviz and pydot (used to save network model)
- h5py (used to save model .h5 file)
- Windows 10
- Python 3.6.8
- CUDA 9.0
- cuDNN 7.4
- tensorflow-gpu 1.9.0
- Keras 2.2.4
- PyQt5 5.15.0
- Nvidia GTX 1060 3G
- Windows 10
- Python 3.7.8
- CUDA 10.0
- cuDNN 7.4
- tensorflow-gpu 1.14.0
- Keras 2.3.1
- PyQt5 5.15.0
- Nvidia MX350 2G
- Raspbian(Debian based)
- etc.