IACC Hackathon 2022 is a nationwide event to provide students with a platform to solve some of the pressing problems we face in our daily lives. Students are invited to choose any problem from the following list.

Poor Water Quality- A Serious Threat

In the beginning of 20th century, Mahatma Gandhi said, “The soul of India lives in its villages”. As per the 2011 census of India 69% of Indians are living in the 640867 different villages with ranging the population from 500 to 10000+ where they have their amenities. In India, there is ministry named as Department of Drinking Water and Sanitation which provides the assistance to the states to provide safe and adequate drinking water to villages with focus on service delivery. However, due to the poor quality of water a large population is get effected by different disease i.e., cholera, diarrhoea, dysentery, hepatitis A, typhoid, and polio etc.

By utilizing the recent technologies i.e., IoT, AI, data science we can provide a solution to regularly check and monitor the quality of water in fixed water tank and based on the report an appropriate action will be taken place.

Essential Knowledge:

1. Web development tools i.e., HTML, Java Script, CSS, SQL server
2. Android/IOS app development tool: Android Studio
3. Raspberry Pi/Arduino Uno
4. Various sensors to check the water quality

Tentative Solution:

Following steps may be followed to complete the above project.

1. Development of IoT module to detect the PH level or any other parameters related to water quality.
2. Collect the data from IoT module and transfer to cloud using wi-fi module on real time.
3. Development of interactive dashboard for display of the data.
4. Development of App for easy access.
5. Based on the collected data, send notification to concern authority for corrective measures.

Conv-Edge: Convolution Neural Network for Edge Devices

The domain of deep learning has paid enormous attention due to its vast application in various domains i.e., Healthcare, Security Surveillances, education etc. Generally, convolution neural network (CNN) is useful for providing the decision on image data. In CNN, data is passed in the input layer and processed by hidden units in intermediate layers. After certain number of epochs, loss is minimum in the network and utilized the model for decision-making purpose. However, it is a challenge for the user to decide the number of layers and hidden units in each layer. Due to excess number of layers and hidden unit the size of trained model is too large and enhances the inference time.

In present time, number of edge devices are increasing exponentially, and these devices can be utilized for decision-making purpose in real time after deployment of deep learning models on these devices. However, it is not easy to deploy these models on edge devices as size of trained model is too large and devices having the limited storage. Moreover, these models are also not properly processed by the edge devices due to limited computational power. Therefore, compression of trained deep learning model is one of the solutions. During the compression process, generally we try to eliminate such nodes or layer which is not providing a good contribution in decision making. However, we have to ensure that the performance of compressed model should not be go down as compare to uncompressed model. In past, many methods such as Matrix factorization, Network slimming, greedy etc., has been developed by the researchers for the compression purpose. However, compression of deep learning model comes under the NP Problem so we can also use meta-heuristic approaches i.e., Genetic Algorithm, Particle Swarm Optimization, Ant Colony Optimization etc. for finding the optimal number of hidden units and layers in a CNN. In this problem, students are supposed to develop a compressed CNN for leaf disease classification.

Essential knowledge:

1. Good in python
2. Good in developing the convolution neural network
3. Strong concepts in deep learning
4. Must have in depth knowledge in any meta-heuristic approach i.e., genetic algorithm

Tentative Solution:

Following steps may be followed to complete the above project.

1. Download the leaf disease dataset from Kaggle (https://www.kaggle.com/datasets/vipoooool/new-plant-diseases-dataset)
2. Train a CNN on leaf disease dataset.
3. Apply the meta-heuristic approach for eliminating the useless hidden layer or unit. For more details about the compression process may read the article at https://link.springer.com/article/10.1007/s12652-022-03793-1.
4. Train the compressed model for a single epoch on same dataset.
5. Test the performance of model ensuring that there is no drop in the performance evaluation metric.

THE EAGLE EYE: Identification of violence activity by individual or mob using drone on streaming video data

In the last two decades, area of deep learning has been paid enormous attention by the research community due to its vast application in various domains such as healthcare, security surveillances, decision making, weather forecasting etc. Moreover, many researchers are also utilized deep learning for human activity recognition inside restricted environment as well as in public place. India is a country where followers of different religions have been living without any conflicts and celebrating festivals in joyful manner. However, during festive seasons different kind of programs i.e., Ramleela, Taj Mahotsava etc. are organized in public places and lot of responsibilities lie with Police & District administration to manage the crowd. In past, it has been noticed that a group of peoples tried to misbehave or indulge in unsocial and unethical activity in crowd which may lead to fight among the groups and responsible persons escaped due to the crowd.

Students are advised to develop a deep learning-based framework for identification of human activity in public as well as restricted environment and if any unusual activity(s) is in existence then model pass the information to respective authority. The size of deep learning model is huge in nature so by using the meta-heuristic approach, first we make it compatible with IoT-enabled camera and deploy for continuous monitoring to the specific region. The collected data is also uploaded on cloud for future reference. Whenever, any non-favourable events are there the intelligent device(s) communicate with the public administration and local police to control the situation without too much loss.

Essential Knowledge:

1. Good in python
2. Good in developing the convolution neural network
3. Strong concepts in deep learning
4. Must have in depth knowledge in any meta-heuristic approach i.e., genetic algorithm

Tentative Solution:

Following steps may be followed to complete the above project.

1. Download the leaf disease dataset from https://ieee-dataport.org/documents/chu-surveillance-violence-detection-dataset
2. Train a deep learning model on the dataset.
3. Apply the meta-heuristic approach for eliminating the useless hidden layer or unit. For more details about the compression process may read the article at https://link.springer.com/article/10.1007/s12652-022-03793-1.
4. Train the compressed model for some epoch on same dataset.
5. Test the performance of model ensuring that there is no drop in the performance evaluation metric.

EduVeri+: A Block-chain based solution for validation and verification of educational degree certificates

Blockchain is a system of recording information in a way that makes it difficult or impossible to change, hack, or cheat the system. Nowadays, blockchain has been paid enormous attention due to its vast use in different areas i.e., CRM, Healthcare, finance, digital currency etc. The major advantage of blockchain is that once information uploaded in the blockchain then it is impossible to change or manipulate the data. Moreover, as blockchain does not allow anyone to manipulate/change/modify the stored information so, we can use blockchain to validate and verify the degree certificates.

In India, 1000+ universities are there and approx. 40 million students are registered in these universities under the various courses. The proposed framework minimizes the burden of govt./private officials who actively involved in the verification of degree certificate demanded by other organization. To minimize the burden of verification, all the data related to degree is uploaded on the blockchain using an interface. When a student successfully completes the course a QR code is generated and this QR code is added in the degree certificate. Anyone can verify and validate the degree certificate by scanning the QR code. As the data is uploaded on blockchain so, there is no chance to manipulate/modify/change in the data. Similar kind of project is on high demand so that we can minimize the cases of fake degree certificate.

Essential Knowledge:

1. Good in programming using Java Script, GO or any programming language which is used in blockchain
2. Good in developing the blockchain
3. Strong concepts in blockchain
4. Must have in depth knowledge in smart contract, web development, hyper ledger.