Smart Surveillance System - S Cubed

Due to the limited attention span, inefficiency, gullibility and questionable ethics, when more humans operate a surveillance system the number of points of failure increases causing a surge in the maintenance cost and a plunging profit margins and the rate of efficiency at which resources are used.

To reduce the point of failures, we introduce the S Cubed AI. As AIs can be active 24/7, their attention is only limited to the hardware we run them on, can't be manipulated by external forces. So, the system will perform in it's most efficient state and will be able to flag most of the significant events across it's data feed.

Our solution can reduce the costs related to maintaining a better Surveillance System at a lower cost and making it accessible to more businesses and home users.

Bangladesh has witnessed a 69.75 % increase in its crime rate in the last 3 years with corruption and bribery being rated at 88.03% by number. With surveillance and identification limited to what it currently is, the imminent advancement in the field of security has become a necessity rather than a choice.

Current security camera pose independent facial recognition inabilities, lacks Suspicious activity detection and  no real time crime analysis and requires high volume employment which eventually fails to identify over half the problems before they initiate. The simple and efficient introduction of the S Cubed AI incorporated with an response method can assess ,analyse and instantly notify government branches, home owners and retail shop owners about customer measurement, demography and persona based targeting through license and facial recognition and foot traffic analysis to the point where employment will only be required for response, eliminating chances of bribery and corruption in the field of security, all while focusing primarily on reducing the cost of integrating such a system into existing surveillance infrastructures to gradually eliminate specific problems in fields of production and security that were once thought to be unsolvable.

S Cubed Corporation is aiming to aid government and non-government agencies primarily working in fields of production and security. With the Aid of the S Cubed AI, we can use facial recognition and license plate identification to flag out illegal activities to the point of reducing their corporate management systemic costs to a minimum. Tolling systems can be automated to allow smoother, faster and a more optimized use of resources in the field of traffic management to allow mass events to occur without causing inconvenience to any surrounding offices of corporate employees. To homeowners we aim to provide a smart monitoring system which can aid them to prevent intruder access and attempted theft, all the while regulating smaller issues like thermal control, to enhance people's way of life. For retail business owners, we can observe and record foot traffic in front of their shop to provide an analysis on customer measurement, demography, persona-based targeting and customer journey time. We aim to integrate the required software into their systems at the lowest possible cost to maximize the shops sales while reducing their management expenses.

S Cubed Corporation plans on releasing sCube, a hardware device and S Cubed AI, a software solution. 

sCube is a smart NVR that has a compute engine (GPU) inside it. S Cubed AI is the software that processes the footage real time and creates a log of persons, vehicles and activities, and notifying the admin or selected users of any suspicious event in the system.

S Cubed AI is essentially the conscience and sCube is the brain of a fully automated surveillance system. It will be able to track vehicle via type, color, license plate, detect humans via face, movement, personal devices (phones, Bluetooth headphones), keep a record of resources being used by the humans and it can then automatically generate analytical data of who and how resources are being utilized. It can be used as an authenticating system, attendance tracking, traffic management system, and a lot more which we haven’t thought about. The sky's the limit here.

Our entire project can be divided into 4 segments:

1. Research and development
   We have currently developed facial recognition, object detection and automatic license plate detection algorithms in-house. We will be developing behavioral analysis, vehicle detection, movement pattern recognition system as well. We have an IoT team who are working on developing WiFi device detection & tracking and phone detection hardware. We plan to integrate all of them into our surveillance solution.
   All of these programs will be added as features to our pipeline, which will take video feeds, images, sensor data, IoT data and process them and output meaningful analytical data which can be stored in a database for further manipulation. We will create the other features based on the data stored on the databases.

2. Data mining
   To train our AIs, we need a lot of data. Primarily, we’ll train our AIs with supervised data collected by us before deployment. But after deployment, the AI will keep learning more about it’s environment using the data collected from it’s feed, which means the AI will need a few days to reach its full potential.

3. Deployment
   We are primarily providing a software solution to businesses and corporations to leverage their existing surveillance infrastructure. If they need processing hardware, we will be recommending off-the-shelf hardware or cloud based on their needs and preferences.
   

4. Further research
   We will be researching on improving and adding more features to our system. Eventually we will come up with more implementations of our system.

S Cubed Corporations is the first to integrate sensor data such as a seismometer, proximity sensor, radar, temperature sensors or any other sensors and developed our own in-house WiFi device detection and Tracking and GSM detection IoT device, which we can deploy to collect more data. This will help our device confirm users with or without video feed. Tailored to the needs of the user the AI can be diverse and configured to be used in any way required at any part of the city.

Also, we are training our AIs to identify people of color, various gender, and other minorities which were previously neglected or sometimes excluded from facial recognition technology. Our License Plate Recognition system is being trained to detect plates in various scenarios, such as dark environment, muddy areas, over exposed image, grainy image, angled or obscured ones. This sets us apart from other solutions.

Our solution will create data which be automatically flagged as suspicious or expected event in a system, thus eliminating the need of a person to report the event. So, every event is equally treated and kept record of.

Having concrete data to back up evidence and being significantly less tedious to extract those data, reporting and the process of taking an action against/for the event will be much easier and faster. Resulting in safer environment, or better for businesses in what form they have deployed our system.

I believe our system will help mitigate the problem of relying solely on human to identify and report events, rather our system will do it much efficiently and effectively at a much lower cost and faster rate.

Currently, we are experimenting with the AI in our own company. 

In the first year, we are targeting to provide services for approximately 10 companies.

In the next 5 years, to provides these services for the Bangladeshi government, 100 companies and deploy the home version of our software on the cloud.

Within the next year, we would like to employ more engineers to release our first version and try to achieve our first milestone of ten companies. 

In the next 5 years, we would like to add more features and improve the existing ones with more research.

The current barriers we are currently facing are:

1. Financial: We need to hire more data scientists and engineers to make an MVP as soon as possible. We need to setup a permanent office location and start a research facility. We need computers, servers and an electronics lab too.
2. Technical: We require technical consultancy to guide us in developing the best product suited for the current market.
3. Legal: We need a legal team to take care of our contracts, business licensing and other legal requirements.
4. Cultural barrier: Ordinary people of Bangladesh are still afraid of technology and very reluctant on accepting new ones.

To overcome the barriers:

1. Financial: We are preparing to ask for Seed Funding or grants to fund our research and recruitment.
2. Technical: We are talking with a few consultants and advisers about the current market and trying to figure about the product ourselves. If we were provided with technical and business consultancy, we'd be much benefited.
3. Legal: We are planning to hire a part-time lawyer to take care of our legal part.
4. Cultural barrier: We want to start a PR stunt such an awareness program or show them in general how our technology will make their life better even without having to interact with it.

We have 5 full time staff, 2 part-time staff and 2 contractors.

We understand the product from both a consumer and an engineer's perspective. We have started working on this project as a software we'd require to automate a few things in our existing company. But later thinking about it's potential, we started working on it as a venture.

Sakib Sadman Shajib, Founder and CEO, who is currently attending a Computer Science & Engineering program in North South University and is working with IDC as an Analyst, is doing his research in Deep Learning and Neural Networks. Noel Alam, BSc in CSE from AIUB, our CTO is serving as our lead developer. Tahsin Ahsan, BBA in Marketing from CUB, Ishraq Hoque, BBA in Entrepreneurship from NSU, Abrar Arafat being in charge of implementing a plan of action to promote and brand the product and with G.M Mehrab Rashid at our helm as COO, thrusting us forward, we expect our product to stand out in a market of several AI assisted services. A few other software engineers are working with us as part-time staff in research and development.

Key resources: 

• In-house facial recognition, ALPR, object and vehicle detection, behavioural analysis and motion pattern recognition algorithms.

• IoT and sensor data analysis program.

• WiFi and GSM detection devices.

• Client Datasets

Partners and stakeholders:

• GPU Manufacturer - NVIDIA, AMD (Prospective)

• Cloud hosting companies - AWS, GCP (Prospective)

• Security Companies (Prospective)

Key activities:

• Writing new AI 

• Training them with newer datasets.

• Research on improving the learning rate and frequency of the results.

Cost structure:

• Cloud hosting - AWS, GCP

• Office - Rent, Furniture, Utility, Employee Costs

• GPU Workstations - A few workstation with a few Tesla V100s or similar

• Computers - 10 computers for office work, development and research

• Networking and Server - A File Server, A VM Server for internal testing, 1GbE/10GbE Networking Firewall, Switch and Router and more.

• Key hires and skills - Data scientists, software engineers.

Type of intervention:

An AI powered smart surveillance system, integrated into the clients existing infrastructure. Our proprietary software will give periodic reports on employee activities, resource utilization, employee behavioral (and also emotional) index, security status and a lot more.

Marketing Channels:

• Internet Ads

• Communities of industrialists 

• Newsletters

Target Market:

• Governments

• Large corporations with 100+ employees

• Home and small office security

Revenue:

• One time license and subscription model.

• Hardware sales and Cloud services.

• Hardware consultancy.

• Licenses or Patents.

Value Proposition:

• Data backed decision making. AI assistant for the HR.

• Better, more efficient and less point of failure security system.

• Better resource utilization.

• Can report about traffic status and suggest better route using AI.

• Report road safety violation.

• Better crowd control feature.

Primarily we plan to apply for funds and grants to cover our pre-deployment expenses. We plan to launch in a fee-for-service model where our clients pay us a subscription or a one time license fee to use our service.

After our first launch, we'll apply for further investments to fund our research and other associated costs.

Tiger IT is a renowned tech company who has access to a network of investors and technical individuals. We applied to Tiger IT Challenge to leverage our business using the network.

We have identified a few key points that'll help us improve our business:

1. Fixed Cost - Employee Salary, Office Rent, Furniture & Utility
2. Variable Cost - Resources for Research (i.e. scientific papers, competitor product or services, electronic lab), Service Marketing, Business Tours.
3. Licenses - APIs, datasets.
4. Hardware - Developer PCs, GPU Workstations, Cloud Subscription, SBCs (NVIDIA Jetson Nano, Raspberry Pi)
5. Key hires/skills - Data Scientists, Software Engineers, Web & App Developers, Business & Legal Advisers, Technical Consultant

Our team will appreciate personal grooming from experts of Engineering Economics, Business Management & Strategic Analysis. This will help us steer our business goal in a orderly fashion.

We'd like to partner with Google, NVIDIA and Microsoft to collaborate on their AI projects. They have some wonderful AI researched and are very willing to collaborate with AI research companies. We'd like to collaborate with them on both closed-source and open-source projects.

We'd also like to work with Gaze Technology, a Bangladeshi AI startup in research too.