Notice: This Project is a subproject of my senior's MSE degree's Paper (palenn (Zhaochangtong) (github.com)). Salute!
My senior's MSE degree paper's content is identifing whether the mutton meatball is fungal-infected and whether the meatball is adulterated with duck by meanings of the fusion of E-nose and NIR(Near Infrared) features.
So I create this repository , try to do the same identification using only the images of those meatballs.
python=3.7 (Anaconda is recommended)
PaddlePadddle 2.4.2(must be 2.4.x ,pd.dataloader()has been refactored by Baidu since 2.5.0)
Install on Windows via PIP-Document-PaddlePaddle Deep Learning Platform
pip install the packages below:
paddlex
PyQt5
scikit-image
jsonpath
filelock
(Just pip install the packages above, nothing else needed)
notice: the token has expired, so the online prediction doesn't work now, the HTTP Reqest Help Document can be found at: https://cloud.baidu.com/doc/EASYDL/s/Sk38n3baq With Baidu EasyDL , the prediction model is trained online. The test image is uploaded to the cloud and EasyDL will return a json file. Just split the json file to exact what you want
FYI: the CPU solution on the right side (“4.2 CPU方案”): SIFT (feature) + Bow(bag of words) + SVM(supporting vector machine) & HOG(Histogram of Oriented Gradient) + Kernel SVM have not been implement
all the pics are shot by a Canon EOS 600D , auto exposure and auto white balance
DataSet:https://pan.baidu.com/s/1UjLCDVn1Yed_4pqet-d26g?pwd=data
the white balance is not accurate by the cam, but we put a white board in the pic as a pure white reference to calculate the white balance afterwards.
the meatball is placed in a round Petri dish, before inputing into the predict algorithm
, I use Hough Circle Transform to detect the round Petri dish and crop it out.
The Filename example : CK-100%-0 (2).JPG or J-45%-2 (5).JPG :
CK / J : CK means no fungal-infection, J means with fungal-infection
0%~100% : means the mutton meatball is 0%~100% adulterated with duck. For instance, 45% means the mutton meatball is 45% adulterated with duck.
I cut 0%~45% as "low" and "60%~100%" as "high"
0~5 : not invoved in this subproject, this number means the meatball were placed in the petri dish for 0~5 days before this photo was shot.