Smart Skin Burn Device
In the past few years Convolutional Neural Network(CNN) is revolutionizing the arena of image analysis. In the following research, we have provided with an effective way of identifying the degree of skin burn and providing with a suitable first aid treatment. We propose a method to recognize b
2025-06-28 16:35:43 - Adil Khan
Smart Skin Burn Device
Project Area of Specialization Artificial IntelligenceProject SummaryIn the past few years Convolutional Neural Network(CNN) is revolutionizing the arena of image analysis. In the following research, we have provided with an effective way of identifying the degree of skin burn and providing with a suitable first aid treatment. We propose a method to recognize burn images through integrating the CNN model. The major aim of this research is to bring a change in the way of treating a burn patient not only in Pakistan but around the world, to step forward in the medical technology, towards automation in identifying the degree of burn through smart device and suggesting a first aid to the patient. The dataset for burn images has been collected by different websites, and we will be collecting the burn images from burn patients at Burns Centre (Civil Hospital), Liaquat Naational Hospital and also Burns Centre at Patel Hospital. The Convolutional Neural Network is adapted and trained to work so the picture taken by the camera installed in Raspberry pi can be identified and treated efficiently according to the degree of burn. The device can be effectively working for industrial sector, health care sector or even for commercial.
Smart devices working upon the methodology of deep learning and convolution neural network is amongst one of the few emerging trends. This project is mainstream for medical sciences and dermatological treatments. This device is made user-friendly and compatible to be used by layman or untrained people. The users of this device can be as many as possible. From students at high schools or professional colleges to working staff at any industry or any ordinary untrained working lady. This versatile project is an innovation in the medical industry. The basis of the working principle and structure of our device it can be fully automated in future.
Project Objectives• To develop a smart device that can detect a burn, identify the degree of burn and provide immediate first aid instructions for its treatment.
• To develop and research methods for identification and classification of burns using deep learning.
• To research and study about the superficial dermal, deep dermal and full thickness type burns.
• To create a cost-effective and user-friendly device for personal and professional use, and to be used in remote areas where medical facilities are not easily accessible.
• To gather and manage a dataset for different degrees of burns and develop the required software using that dataset.
• To set up Raspberry Pi for deep learning and training the network
We look forward in making a Raspberry pi plus Neural Network based smart device. The main and basic purpose of this device is to take pictures of the burn manually and then according to the severity of the burn it prescribes us the first aid treatment of the affected area (Burnt area).
The Hardware of this project includes
- Raspberry pi
- Camera
- LCD
While the Programming language is
- Python
The reason for choosing Raspberry pi as the microprocessor of our project is that, it contains a built-in port for Camera and LCD which is attached as hardware to the microprocessor. Moreover, Raspberry pi is Linux based which means that it has built in python installed in it. The python based Convolutional Neural network and Deep Learning code is installed to our Raspberry pi.
When our device is once ready which clearly means that it is trained completely according to the data set and the Convolutional Neural Networking and Deep learning code is installed in Raspberry pi, then we manually take pictures of the burnt area through camera after which our code runs to identify the picture through Convolutional Neural Networking which is a biologically-inspired programming paradigm which enables a computer to learn from observational data according to what it is trained previously and Deep learning which is a powerful set of techniques for learning in neural networks. It identifies the picture to the given Degrees of burn and after this, it treats the burn accordingly. There are approximately 2% chances that somehow our project does not identifies the picture taken from the patient due to common issues like unclear images or blurred vision or not visible enough.
PRELIMINARY RESULTS/PROOF OF CONCEPT:
The screenshot of our application below provides us with the proof of what we have done till now. There are 4 buttons provided below:
- Camera: which takes Live Images for identification and prescription of the burn.
- Select image: this is an additional function. It can be used to identify any image present in our laptop earlier/in the dataset
- Show result: this displays us with the degree of burn and the prescribed first aid treatment
- User guide: this is for guiding any unknown person about how to use this device.
Beneficiaries
- It eliminates the need of going to a doctor for less severe burns.
- It gives first aid instruction for in-home treatment.
- It enables the user to accurately treat different degrees of burns until further assistance is provided, hence eliminating the severe consequences that result from mistreatment of burns due to lack of knowledge/expertise.
Moreover, using the same technique that have been used for burns, this device can be further enhanced to work on detection of different skin diseases in the future so it has a very wide scope.
Technical Details of Final DeliverableThe processing has been done through Transfer Learning which is a Deep Learning technique. The algorithm that we have used in transfer learning is called Inception V3. We have trained our model through this algorithm. The reason we have used Inception V3 to train our model is that this is the most accurate algorithm in terms of image data apart from yolo.
To make the application user friendly and more interactive, we have added five buttons in the menu. One of them is camera that will directly open the camera and allow users to take an image at any instant. Another option is select image which can be used if the user wants to recall an image taken in the past and observe what results were generated. The show result option generates the result of the current image being processed and burn specialist button will give a basic introduction of what the application does. Lastly, the user guide button tells the users how the application can be used.
After the completion of application, we will install our Software in our Smart Device which is Raspberry pi based. We have used Raspberry Pi (4 Model B) and interfaced it LCD (7inch IPS touch screen for raspberry pi 4 B) and Camera (for RP 4B model)
The reason behind using Edge TPU accelerator is to speed up the processing of the image through Transfer Learning technique since it is time consuming, and to increase the capacity of Raspberry pi i.e. the trained model and the dataset is consuming a lot of space in Raspberry pi so to increase the speed we trade off cost of the TPU accelerator.
The final outcome of our device will be identifying the degree of burn (up to 4) with its first aid treatment and medicines plus some preventative measures that must be immediately taken.
It is hereby noted that these will be authentic medicines and prescriptions that will be provided by the mentioned well-known hospitals
• Burns Centre at Civil Hospital
• Liaquat National Hospital
• Burns Centre at Patel Hospital
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 47000 | |||
| Raspberry Pi (4 Model B) | Equipment | 1 | 9500 | 9500 |
| LCD (7inch IPS touch screen for raspberry pi 4 B) | Equipment | 1 | 9000 | 9000 |
| Camera (for RP 4B model) | Equipment | 1 | 4500 | 4500 |
| Edge TPU accelerator | Equipment | 1 | 15000 | 15000 |
| Other Expenses(Including cases, boards, wirings and etc) | Miscellaneous | 1 | 9000 | 9000 |