Model Advantages

Some of the advantages of the Learning Model are:

• Tensors are used for faster processing
• Transfer Learning gives us the advantage of using the well-known pre-trained models which can help improve the performance on the dataset
• Due to transfer learning, the used models have been pre-trained on a general data set and needs to be trained only for the specific characteristics and since the models are deep neural networks, so it can train easily even on a smaller number of epochs which can reduce the amount of computation, processing and use of resources.
• Ensemble model is used to help improve the overall accuracy of the model by bringing together several models as one and create a new deeper architecture.
• In this project, the image file of the patient is upload into a public use software, which is a GUI-based interface, developed with the help of Tkinter and Python, and it consists of the model and the software processes the image through the model and predict the results which can help people and doctors to start with the medication way earlier instead of waiting for the laboratory tests and reports for the confirmation.

Objective

Skin cancer is an alarming disease for mankind. The necessity of early diagnosis of the skin cancer have been increased because of the rapid growth rate of Melanoma skin cancer, its high treatment costs, and death rate. This cancer cells are detected manually, and it takes time to cure in most of the cases. We even intend to help people in the pandemic time so that people do not have to risk their lives on getting the laboratory tests.

We want to build an application which we can deploy in the production for the public and can be used by anyone and everyone for their benefit to start early with the medication and precautions.

Motivation

Skin cancer is an abnormal growth of skin cells. Most skin cancers are caused by exposure to ultraviolet (UV) light. When the skin is not protected, UV rays from sunlight or tanning beds can damage and alter skin's DNA that leads to the cancer.

Deep learning model has been built to classify and identify the binary diagnostic group of melanocytic images obtained through dermoscopic. Based on the model, disease detection through dermal cell images has been investigated, and classifications on dermal cell images have been performed.

Background

Skin cancer is the most prevalent type of cancer. Melanoma, specifically, is responsible for 75% of skin cancer deaths, despite being the least common skin cancer. The American Cancer Society estimates over 100,000 new melanoma cases will be diagnosed in 2020 and almost 7,000 people will die from the disease. As with other cancers, early and accurate detection—potentially aided by data science—can make treatment more effective [1].

Dermatologists could enhance their diagnostic accuracy if detection algorithms consider “contextual” images with the same patient to predict which images likely melanoma. If classifiers would be more accurate could better help dermatological. The ISIC Archive contains the largest publicly available collection of quality-controlled dermoscopic images of skin lesions. You will use images and predict the likely of melanoma. Using patient-level contextual information may help the development of image analysis tools [2].

Melanoma is a deadly disease, but if caught early, most melanomas can be cured with minor surgery. Image analysis tools that automate the diagnosis of melanoma will improve dermatologists' diagnostic accuracy. Better detection of melanoma can positively impact millions of people [3].

Executive Summary: Skin Cancer

Skin Cancer is categorized into two categories namely: malignant melanoma(lethal) and benign(non-lethal/treatable). Evolution in technology has brought upon many boon’s as well as banes and Cancer is one of them, skin cancer becoming a common side effect due to exposure to a lethal level of radiation or due to genetic disorder from exposure to radiation. The opportunity for machine learning models to be implemented, recommendations and costs are mentioned in this article.

Opportunity

With boons of advancement in technology machine learning models can be implemented and trained to detect, diagnose and possibly prevent any benign case from turning into a malignant case thus, advancing the medical infrastructure. Currently the main challenges are the cost-effective models to train, economic viability and error detection rates varying from ethnicity/race and a few specific cases.

Solution Recommendation

Training of a model with Mobilenet, Inception and Xception Architecture gives an economically viable solution and a detection rate of above 60% which as per the IEEE standards is viable for use practically. Recurring training of various Skin Cancer training data based on specific individuals with previously known other skin disorders and others on their race/ethnicity as to increase the detection rate without overfitting the model.

Skin Cancer Rates: Both Genders

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Skin Cancer Rates: In Men

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Skin Cancer Rates: In Women

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Model Development and Training

Development of the Machine Learning/Deep Learning Model requires the knowledge and skill of using Jupyter Notebook (either installed via Anaconda Studio or directly as an individual component) or Google Colab along with high-powered Graphical Processing Unit for better, reduced and efficient processing time.

Use of Google Colab gives more advantages for writing and execution of code in Python, sharing notebook, Integrating Libraries and Dependencies and also to increase the available processing power through Cloud Computing. The advantage of Cloud Computing allows the model to still train even if the system crashes due to some unforeseen reason and does not cause a halt to the training.

Interface Development and Testing

The interface is converted into an executable file (exe File) so that the general public regardless of with or without any technical skills can simply install the software on their system and use the model to check if the patient suffering from the disease has Skin Cancer or not, so that the patient can think and take action at an early stage and start with the precautions and medications which can help save his/her life.

After uploading an image of the infected area, the final result is that whether the person is suffering from skin cancer or not. The final result is shown through an alert box.

Interface Requirements

There are two interfaces namely – User Interface and Software Interface:

• User Interface: The user interface will be implemented using any desktop running on Windows OS. This interface will be very user friendly so that people from different strata can use it to detect their disease without any difficulty by just uploading their medical test image.
• Software Interface: A software interface running on Windows OS. It should have Python compiler.

Platform, Configuration and Specification

Platform: Google Colab (Pro version recommended)

Standard RAM: 12.69 GB

High-RAM Configuration: 25.46 GB (recommended with Pro version)

Normal Disk Space: 89.15 GB

Pro Version Disk Space: 147.15 GB

Engine: Google 3 Python Compute Engine backend (GPU)

GPU: NVIDIA Tesla K80

Design Approach/Materials and Methods

As suggested by the mentor to use the pre-trained models, for example, VGG, ResNet, MobileNet, etc., we planned on using these models and building an architecture. The proposed model includes the use of MobileNet, Inception and Xception model which are very well-known pre-trained architectures. These architectures have been brought together using Ensemble Modelling in which we have created a list of these models through which we join the Neural Net and further join their weights for one single model and to used and updated.

Code and Standards

For the implementation of this model, the code was written using the programming language Python. Python also has several advantages due to its hundreds of open-source libraries which are available for almost every functionality to improve and increase the performance. The program for interface is also written using Python.

Source Code in the form of a PY File

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Source Code Screenshots in Colab

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Pre-processing

All the required libraries are first loaded into the model and then the dataset is also loaded into the model using the Google Drive as the Dataset has been uploaded to the Google Drive it can be accessed directly through it.

After the Google Drive is connected and authorised for use, the CSV files of the Training, Validation and Testing Data are read one-by-one and the data which they contain is displayed. Further, it is required for the dataset images to have their directory location attached to the filenames because then the images can be accessed directly with it without changing the directory which might interrupt any further operation.

After the directory locations are attached to the filenames, the images are loaded into the model one-by-one and they are converted into tensors. After obtaining the tensors, they are stored in a separate location so that these tensors can be used directly and it eliminates the need for loading the dataset and processing it again and again and instead these tensor files can just be loaded into the model and used for further training and processing. This also saves a lot of time because we do not need to create the tensors again and again as the tensors will be same each time they are being created because it is the same images which might be used during the development of the project.

Training Model

For this, we have used Transfer Learning and Ensemble Modelling. For the Transfer Learning, we have used the following pre-trained models:

• MobileNet Architecture
• InceptionV3 Architecture
• Xception Architecture

For the training, the pre-trained models are first loaded in as a function and then a since we have a checkpoint of these models, so we can start training these models from that checkpoint again. We start training it further by fitting it on the data which we have using the tensors.

Prediction

This prediction is done on a sample image just to check that all the models are working and are predicting. Working of this section denotes that the model can be taken to the further stage of evaluation of the validation data and the testing data.

For this part, we take the sample image and convert that image into a tensor. After that, the Individual Models and their respective weights are loaded so that the prediction can be made using the predictive function which we have defined. This is the function which will also be used for the Interface Application for the prediction on the image which will be uploaded by the user.

Evaluating the Models Individually on Validation Data

For the evaluation of the models on the Validation Data, we have used the Receiving Operating Characteristic Curve. A receiver operating characteristic curve, or ROC curve, is a graphical plot that illustrates the diagnostic ability of a binary classifier system as its discrimination threshold is varied. The method was originally developed for operators of military radar receivers starting in 1941, which led to its name.

For this, the we will be computing the test set predictions and then the model will be evaluated based on the Receiving Operating Characteristic Curve whose values will be obtained using the Confusion Matrix. Then, we will plot the values obtained for this in order to obtain the graph and check for the best possible values which can be used for prediction.

Further, after obtaining all these values, we will be calculating the Precision, Recall, Specificity, Sensitivity, Negative Predictive Value and Accuracy for further evaluation and after this, we can check the parameters, and if possible, we can tune the parameters to obtain better results.

At the end, we can compare the test results for the pre-trained architecture, as we have noted that the loss had reduced to the most efficient values possible and if we will train the model further with more epochs on the architectures, then we might make the model overfit on the data which can decrease the results and might lead to a higher number of wrong predictions, furthermore reducing the accuracy of the model and its efficiency.

Evaluating the Models Together on Validation Data - Ensembling the Models

Defining a specific input shape for all the input images. The models will be brought together and appended so that the image passes through all the models one-by-one. The ensemble model will be defined which will contain all the models and the then the weights of the model will be combined and a file will be generated so that the weights are together into one and then they can be used together for the for the complete architecture.

For evaluating the models, we will be using the test set predictions and the ROC curve will be used whose computational values will be obtained using the Confusion Matrix. The other evaluation metrics are Precision, Recall, Sensitivity, Specificity, Negative Predictive Value and Accuracy. The results obtained for the Ensemble Model are better than each architecture individually.

Evaluating the Models Individually on Testing Data

For the evaluation of the models on the Testing Data, we have used the Receiving Operating Characteristic Curve. A receiver operating characteristic curve, or ROC curve, is a graphical plot that illustrates the diagnostic ability of a binary classifier system as its discrimination threshold is varied. The method was originally developed for operators of military radar receivers starting in 1941, which led to its name.

For this, the we will be computing the test set predictions and then the model will be evaluated based on the Receiving Operating Characteristic Curve whose values will be obtained using the Confusion Matrix. Then, we will plot the values obtained for this in order to obtain the graph and check for the best possible values which can be used for prediction.

Further, after obtaining all these values, we will be calculating the Precision, Recall, Specificity, Sensitivity, Negative Predictive Value and Accuracy for further evaluation and after this, we can check the parameters, and if possible, we can tune the parameters to obtain better results.

At the end, we can compare the test results for the pre-trained architecture, as we have noted that the loss had reduced to the most efficient values possible and if we will train the model further with more epochs on the architectures, then we might make the model overfit on the data which can decrease the results and might lead to a higher number of wrong predictions, furthermore reducing the accuracy of the model and its efficiency.

The results obtained on the Testing data are very similar to the results obtained on the Validation data.

Evaluating the Models Together on Testing Data - Ensembling the Models

Defining a specific input shape for all the input images. The models will be brought together and appended so that the image passes through all the models one-by-one. The ensemble model will be defined which will contain all the models and the then the weights of the model will be combined and a file will be generated so that the weights are together into one and then they can be used together for the for the complete architecture.

For evaluating the models, we will be using the test set predictions and the ROC curve will be used whose computational values will be obtained using the Confusion Matrix. The other evaluation metrics are Precision, Recall, Sensitivity, Specificity, Negative Predictive Value and Accuracy. The results obtained for the Ensemble Model are better than each architecture individually.

Localization

For this, we define the Class Activation Map and then we create another function using which we can display the results obtained on a particular file, as we can use the address of the that image directly and then plot to even check the results obtained for the model which we have created. The results are displayed along with the plot which is created containing the Image and another heatmap image displayed along.

Saving the Complete Model for the Python Interface Application

First, we will move to the required directory where we can save the file for the model. Then, we will use the complete model and use the save() function.

Converting the .h5 File to .tflite File for the Python Interface Application

We will first load the saved model file and then we will use the converter function to convert the file into the into the required format which we can use for the Interface Application.

Data Selection

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Background

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Exploratory Data Analysis

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Data Pre-processing

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Feature Transformation

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Model Selection

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Model Training

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Application Interface

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Mounting Google Drive

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Loading and Display the CSV Dataset

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Converting the Data into Tensors

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Fitting into the Models

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Predictions on a Sample Image

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Evaluation of MobileNet on Validation Data

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Evaluation of Inception on Validation Data

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Evaluation of Xception on Validation Data

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Evaluation of Models Together on Validation Data

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Evaluation of Models Individually on Testing Data

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Evaluating the Models Together on Testing Data - Ensembling the Models

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Localization

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Saving the Complete Model for Python Interface Application

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Converting the .h5 File to .tflite File for the Python Interface Application

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Websites

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Concept & Information

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.

Research

Lorem ipsum dolor sit amet, consectetuer adipiscing elit. Aenean commodo ligula eget dolor. Aenean massa. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Donec quam felis, ultricies nec, pellentesque eu, pretium quis, sem. Nulla consequat massa quis enim. Donec pede justo, fringilla vel, aliquet nec, vulputate eget, arcu. In enim justo, rhoncus ut, imperdiet a, venenatis vitae, justo. Nullam dictum felis eu pede mollis pretium. Integer tincidunt. Cras dapibus.