0312 502 05 42 | ARPANET

Address: ODTÜ Bilim İnavasyon Merkezi – M.Kemal Mah.Dumlupınar Bulv.No:280/G Çankaya Ankara Türkiye

SPOOK | Facial Recognition And Image Processing

It provides a high-performance, low-hardware enterprise facial recognition solution that supports face detection, tracking of the detected face, extraction of the mathematical value of the face, and comparison with the face pool with multiple algorithms.

We’re bringing you the latest and most advanced facial recognition technology, the most effective solution for secure, rapid authentication and person searches, which uses artificial intelligence and image processing technology to scan users’ faces to verify their identity and capture similar images.

Spook brings together four functionalities that play an important role in the field of image processing!

  1. Image classification: Determines whether an image given as input belongs to a particular class and labels the correct class.
  1. Object detection in image: Measures the similarity between two images and scores them on a scale. Provides an underlying framework for determining the ratio of similarity between different images.
  1. Object similarity: Identifies objects in an image given as input, defines these objects, and marks their locations.
  1. Image segmentation: Allows an image given as input to be broken down into specific regions or objects, making it easier to analyze the image.

How it Works?

Spook uses image processing technology and artificial intelligence to analyze image properties. This software identifies the properties of any image content and compares these properties with the recorded data, presenting the results of the analysis.

For facial recognition, for example, the shape, size, texture, position of the eyes, the shape of the nose, the structure of the mouth, and other similar qualities of the face are analyzed by multiple facial recognition algorithms to create a unique facial profile of a person.

Spook Technology

Spook provides you with solutions in a field that has been growing rapidly and proliferating in recent years, and these technologies used to recognize, authenticate and classify faces of people play a critical role for many applications in different industries.

Facial recognition technologies based on Artificial Intelligence (AI) have developed rapidly thanks to recent technological advances in areas such as image processing, data mining, and machine learning, which have increased the efficiency, accuracy, and reliability of facial recognition systems.

Especially in recent years, facial recognition technologies based on Deep Learning have begun to yield better results than previous facial recognition systems.


The Spook project can provide many benefits in places such as government institutions, official agencies, and ministries. Firstly, Spook’s high performance and low hardware requirements can help government and official institutions save on budget and resources. Additionally, Spook’s facial recognition technology is an important security tool for governments and official institutions due to being the most effective solution for secure identity authentication and person investigations.

Spook can also bring significant benefits to police forces and security agencies. It can be used for security cameras and security systems. This can assist security forces in solving crimes and finding criminals. Additionally, facial recognition technology can be used to control entries and exits to secure areas. For example, cameras monitoring passage at a building entrance can be used for purposes such as tracking people or animals, or tracking a specific object (such as a weapon).

Furthermore, Spook can be utilized in the finance and banking sector. Banks and financial companies can expedite customer identity verification using facial recognition technology. This can help customers conduct transactions more quickly and securely. Spook can also be employed in the healthcare sector. It can be used for patient tracking, recording treatment data, and managing patient records.

In educational institutions, Spook can be used for student tracking. Exam processes can be managed using facial recognition technology. In the field of product identification and labeling, Spook can be used to identify and label product photos. Additionally, Spook can be used for the analysis and evaluation of advertisements. Analysis of visual elements used to ensure advertisements reach a specific target audience can be conducted. It can also be used for traffic analysis. For example, it can be used to determine traffic density, track the number and types of vehicles, and collect necessary data for urban planning purposes.

Advertising Management
Spook technology can enhance the accuracy and effectiveness of social media advertisements. Ads can be shown only to the target audience you’re interested in.
Trend Analysis
Spook technology can help identify social media trends and popular topics. It produces solutions for market research and identifying products/services.
Social Media Data Analysis
Spook image classification and similarity functions can be used for analyzing social media data. For instance, it can be used for evaluating the effectiveness of a brand’s advertising campaign, identifying popular visual content on social media, or tracking the similarity of photos belonging to a user’s profile.

Finance and Banking
Banks and financial companies can expedite customer identity verification using facial recognition technology.

Spook can be used in the healthcare sector for patient tracking, recording treatment data, and managing patient records.

Schools and universities can track students using facial recognition technology. Educational institutions can manage exam processes using facial recognition technology.

Product Identification and Labeling
Spook can be used for object detection and image classification functions, to identify and label product photos.

Image Verification and Validation
Spook image similarity functions can be used to evaluate the accuracy and validity of images. For example, to ensure that product images on an e-commerce site are accurate and up-to-date, similarity between product images and those provided by the manufacturer can be checked.

Advertising Analysis
Spook image classification and object detection functions can be used for analyzing and evaluating advertisements. For example, analysis of visual elements used to ensure advertisements reach a specific target audience can be conducted.

Traffic Analysis

Spook’s image object detection and segmentation functions can be utilized for traffic analysis. For example, it can be used for purposes such as determining traffic density, tracking the number and types of vehicles, and collecting necessary data for urban planning.


Efficient Processes

The end-to-end image processing process is shaped with the principle of low hardware requirement and high performance.

Advanced API

Spook offers an advanced and simple API infrastructure. You can integrate it into your existing projects or build your own applications.


It can work on a closed network without the need for the internet. Thus, it allows you to use it securely on your local network.

High Performance

The entire system is developed with a low-level programming language that supports MULTITHREAD, thus providing high performance with low hardware requirements.

CPU – GPU Support

Adjustable CPU and GPU support is available for face detection, mathematical value generation, and comparison operations.

512 Reference Points

At least 512 reference points are taken for face detection and mathematical value generation. Compression is applied to create a 2-kilobyte value to reduce hardware costs.

Distributed Architecture

The platform can work in a distributed and centralized architecture. Thus, it avoids the large network traffic generated by image streams.

CNN Architecture

The most important components of CNN are Convolutional, Pooling, and Fully Connected (Dense) layers. The CNN architecture in our project is optimized for customizing, discovering, and classifying image data. Parameters such as the number of layers used, filter sizes, and pooling operations are optimized according to the characteristics of the image data.

Image Similarity

In our project, SNN is used to measure image similarity. Two image data are customized by the same network structure, resulting in a similarity score. However, Triplet Loss Network and Contrastive Loss Network with different structures and loss functions are used to achieve higher success.

Object Detection (YOLO)

YOLO (You Only Look Once) dataset size is kept large to improve performance and allow the model to recognize more types of objects. The number of convolutional layers in YOLO’s structure has been increased, and larger filter sizes have been used. Overfitting situations of the model are prevented using regularization parameters and other hyperparameters in the SGD optimization algorithm.

Secure Data Support

All data located in both the central and distributed terminals are kept up-to-date and securely in a NoSQL database. All data transfers are encrypted with 512 AES, except for SSL.

Album Support

It provides support for having multiple images for both individuals and objects. Thus, you can create an album for a person and perform facial recognition from dozens of images under it.

Operating System

All software belonging to the platform operates on licensed operating systems such as GPL, GNU (General Public License). Thus, you can save on server operating system costs.

Terminal Logic

Images transmitted to the center via IP and/or web cameras are processed location-based due to the high bandwidth requirement it will cause. Comparison and cropped detection image processing are carried out centrally.

Set the Threshold Value

You can manage the threshold value according to the features of the terminal via a very simple settings file.


The algorithms used in our project have a higher accuracy rate compared to our competitors.

Our project has a faster processing speed compared to our competitors.

Our project can work with a greater number of datasets compared to our competitors and may have more data support.

Our project utilizes more advanced detection techniques compared to our competitors.

Our project consumes less power and requires fewer hardware resources compared to our competitors.


What is the main objective of the project?

The main objective of the project is to perform facial recognition processes and similar underlying processes such as image classification, image similarity, image object detection, and image segmentation.

Is the project user-friendly?

The project has a user-friendly interface and is easy to use. However, some operations may require some technical knowledge.

What is facial recognition software used for?

Facial recognition software allows for the recognition, classification, or identification of faces in images captured from a camera or image source. It finds applications in various sectors including security and surveillance systems, biometric entry systems, advertising and marketing analytics, social media, and photo analytics. Facial recognition technology can be used to verify identities, enhance security, collect and analyze data.

What is the accuracy rate of facial recognition software?

The accuracy rate of facial recognition software is the percentage of accurately predicting that the face identified is indeed that of the person. The accuracy rate may vary depending on factors such as the size of the dataset, diversity of training data, quality of the image source, among others. Generally, under correct conditions, Spook offers a 99.2% accuracy rate. However, proper configuration is important.

How does facial recognition software protect personal data?

Spook takes several measures to protect personal data:

Data Encryption: Encrypting the data collected and stored by facial recognition software enhances data security.

Authorization and Access Control: Only authorized users are allowed access to the data.

Use of Updates and Firewall: The software is regularly updated and firewall is used for security purposes.

Compliance with GDPR and other Personal Data Regulations: Spook is compliant with the General Data Protection Regulation (GDPR) of the European Union and other personal data regulations.

Data Deletion or Anonymization: Spook ensures that unused data is deleted or anonymized over time.

What can your project’s image classification functions do?

Our project’s image classification functions can:

Identify and categorize objects or scenes in an image based on predefined classes. Use neural network models developed through analysis and training of previous data to classify objects or scenes in an image. Use image preprocessing and data augmentation techniques to increase accuracy rate. Assign images to specific classes (e.g., house, car, animal, etc.). Identify and classify objects in images. Analyze images based on the objects within them to perform classification.

What is the legal framework related to facial recognition software?

The legal framework related to facial recognition software varies from country to country. Some countries allow the use of facial recognition technology with limited regulations, while in others it is prohibited. Additionally, the ways in which facial recognition data can be collected, stored, processed, and shared may be restricted by legal regulations. Spook provides technical infrastructure and software. The responsibility for data privacy, usage method, and other matters related to facial recognition software lies with the users. It is important for these users to ensure compliance with legal regulations and protection of personal data.

How do your image similarity functions work?

Our project’s image similarity functions involve structures used to measure the similarity between two input images. These structures typically involve extracting, transforming, and comparing features of the images to calculate similarity. Image similarity functions can be used in applications such as comparing image content, searching for similar images in a picture database, detecting differences or matches between images.

What types of data are supported by your image object detection functions?

Our project’s image object detection functions support RGB images. These can be applied to image data in standard photo files and aim to predict the positions of objects. Image data is first processed and then classified by artificial neural networks, and the positions of objects are determined.

How is image segmentation performed and what is its purpose?

Image segmentation is an image processing technique used to identify different regions in an image and label these regions with different colors or tags. This technique plays an important role in applications such as identifying, classifying, and separating objects in an image. Image segmentation involves steps such as dividing the image, applying learning and analytical techniques.

What platforms or systems is your project compatible with?

Generally, our applications are compatible with operating systems such as Windows, MacOS, and Linux, and use the C++ programming language. They may also be compatible with mobile platforms, depending on the specific requirements of the project and the platforms targeted.

What are the hardware and software requirements for running the project?

The hardware and software requirements for running the project may vary depending on the specific project or intended use. Generally, good GPU and high-memory computers are required for image processing and learning models to work.

Are there training and support mechanisms for using and configuring the project?

Training and support mechanisms for using and configuring the project can be provided. Different options such as online or offline training materials, video tutorials, classes, or workshops may be offered. Additionally, other support mechanisms such as online forums, documentation, or customer support services may be available. These mechanisms will help users understand and use the project effectively.

What guarantees are provided regarding the security and privacy of the project?

Security and privacy are important issues, and care is taken to address them. Current security measures and applications are used to ensure the security and privacy of the data. Additionally, access permissions to user data are controlled, and necessary security measures such as firewall and encryption technologies are applied.

Can you provide information about the project’s development process and future plans?

The project undergoes continuous updates and improvements for the highest performance and user experience possible. In the future, more advanced classification and similarity analysis functions supported by additional features and datasets may be added. Additionally, it is made compatible with different platforms and systems for ease of use and accessibility.

Can you provide detailed information about the working principles and algorithms of the project?

The project’s purpose is to identify and classify specific objects or characteristics in images provided as input. Among the algorithms used for this purpose are methods such as Convolutional Neural Networks (CNN), Siamese Neural Networks (SNN), YOLO, etc. Factors such as optimization algorithms used to improve the project’s performance, loss functions used in training and testing data, etc., are also important. If more detailed information about the project’s working principles is required, please contact us.

How is the update and evolution process of the project conducted and at what frequency?

The update and evolution process of the project is determined based on factors such as customer and user requests, technological advancements, and addressing functionality gaps.

What is the price of the project?

The price of the project can vary depending on factors such as the intended use of the product, its size, and customization options. Please contact the support team to inquire about the price of the product.


The superiority of a neural network algorithm can vary based on factors such as performance, accuracy, speed, design, reliance on training data, model quality, amount of required information, adaptability, ease of updating and development, versatility of use, and similar factors.

For example, CNN (Convolutional Neural Network) is a popular choice for image data and excels in tasks such as object recognition or classification within images. Additionally, other algorithms like YOLO (You Only Look Once) or RetinaNet also yield good results in object detection tasks.

Another example is the Siamese Neural Network (SNN), which can be used for image similarity tasks and is specifically designed for this purpose. SNN is an algorithm used to compute the similarity between two images and may be more efficient and faster than other algorithms for this task.