pyCONVCLASSIFIER

Version: 0.0.5    
Author : Md. Nazmuddoha Ansary, Shakir Hossain

Version and Requirements

numpy==1.16.4
tensorflow==1.13.1
Keras==2.3.1
Python == 3.6.8

Create a Virtualenv and pip3 install -r requirements.txt

Preprocessing

Check The following Values in config.json

  {
      "FLAGS":
      {
          "DATA_DIR"        : "/home/ansary/RESEARCH/MEDICAL/Data/",
          "OUTPUT_DIR"      : "/home/ansary/RESEARCH/MEDICAL/Data/",
          "CLASSES"         : "eczema,psoriasis",
          "IMAGE_DIM"       : 64,
          "ROT_ANGLES"      : "5,10,15,20,30,45,60,75",
          "EVAL_SPLIT"      : 0.05,
          "TEST_SPLIT"      : 0.2,
          "BATCH_SIZE"      : 128
      }
      
  }

DATA_DIR should have the following folder tree:

          Data
          ├── eczema
          └── psoriasis

run main.py

Results

If execution is successful a folder called OUTPUT_DIR/DataSet should be created with the following folder tree:

  ├── Images
  │   ├── {Xid}_{xrot}_{xflip}_{xsample}_{xlabel}.png
  │   ├── {Xid}_{xrot}_{xflip}_{xsample}_{xlabel}.png
  ------------------------------------
  ------------------------------------
  │   ├── {Xid}_{xrot}_{xflip}_{xsample}_{xlabel}.png
  │   └── {Xid}_{xrot}_{xflip}_{xsample}_{xlabel}.png
  ├── Test
  │   ├── X_test.h5
  │   └── Y_test.h5
  └── Train
      ├── X_eval.h5
      ├── X_train.h5
      ├── Y_eval.h5
      └── Y_train.h5

Train Data Size:24192 images
Test Data Size: 6272 images
Eval Data Size: 1152 images

ENVIRONMENT

OS          : Ubuntu 18.04.3 LTS (64-bit) Bionic Beaver        
Memory      : 7.7 GiB  
Processor   : Intel® Core™ i5-8250U CPU @ 1.60GHz × 8    
Graphics    : Intel® UHD Graphics 620 (Kabylake GT2)  
Gnome       : 3.28.2

TPU(Tensor Processing Unit)

TPU’s have been recently added to the Google Colab portfolio making it even more attractive for quick-and-dirty machine learning projects when your own local processing units are just not fast enough. While the Tesla K80 available in Google Colab delivers respectable 1.87 TFlops and has 12GB RAM, the TPUv2 available from within Google Colab comes with a whopping 180 TFlops, give or take. It also comes with 64 GB High Bandwidth Memory (HBM). Visit This For More Info

Training

MODEL: CONV_NET

in convNet.ipynb : SECTION: ConvNet Model Training SUB-SECTION: Data, change the values of the following params as needed:

Model Structre
run convNet.ipynb in colab

F1 SCORE: 98.4375 %

MODEL: DenseNet

The model is based on the original paper:Densely Connected Convolutional Networks

Authors and Researchers: Gao Huang ; Zhuang Liu ; Laurens van der Maaten ; Kilian Q. Weinberger

The paper introduces Dense Blocks within the traditional convolutional neural network architechture.
The composite layers can also contain bottoleneck layers

As compared to well established CNN models (like : FractNet or ResNet) DenseNet has:
* Less number of feature vector
* Low information bottoleneck
* Better Handling Of the vanishing gradient problem

Default Model Structre
In addition to the previous parameters like convNet, the following valuese may be changed for exploring.

run denseNet.ipynb in colab

F1 SCORE: 99.1390306122449 %

Deployment

run app.py
used model: DenseNet
After Training in colab,download the weights from google drive and place them in models folder in the local copy of the repo

mnansary/pySKIND