Understanding Deep Learning: A Comprehensive Guide for Tech Enthusiasts

Ashish Kasama|January 1, 2024|30 Minute read|
Blog / Technology Posts / Understanding Deep Learning: A Comprehensive Guide for Tech Enthusiasts




Deep learning is an advanced subset of machine learning, a method of data analysis that automates analytical model building. It's based on the idea that systems can learn from data, identify patterns, and make decisions with minimal human intervention. Deep learning structures algorithms in layers to create an "artificial neural network" that can learn and make intelligent decisions on its own. This neural network is designed to mimic the human brain, albeit far from matching its ability, and consists of thousands or even millions of simple processing nodes that are densely interconnected.

The concept of deep learning is not new, having been around since the 1940s, but it has gained immense popularity in recent years. This resurgence is largely due to the availability of massive amounts of data (big data) and substantial increases in computing power. Early neural networks were shallow because they didn't have enough processing power to build large networks. However, the advent of the internet and the digitalization of society resulted in an explosion of data and, subsequently, more complex and deeper networks.

Deep learning's importance in today's world cannot be overstated. It's at the core of many technologies that we use daily, from facial recognition in smartphones to language translation services. It powers the most advanced autonomous vehicles, enabling them to recognize a stop sign, or distinguish a pedestrian from a lamppost. It's revolutionizing healthcare by assisting in diagnosing diseases and personalizing treatments. In the business world, it's used for fraud detection, customer relationship management, and many other applications.

The significance of deep learning lies in its ability to process and learn from enormous amounts of unstructured data such as texts, images, and videos. Its ability to continuously learn and improve from experience makes it a valuable tool in the quest to solve many of society's most challenging problems and to drive innovation across industries. As we generate more data and as computing power continues to grow, deep learning will become even more pervasive, shaping the future of technology and society.

Basics of Deep Learning

Deep learning, a subset of machine learning, is a powerful tool that has revolutionized many fields, from computer vision to natural language processing. At its core are neural networks, complex structures inspired by the human brain, designed to recognize patterns and make decisions.

Neural Networks: The Building Blocks

Neural networks are the foundation of deep learning. They are composed of nodes, or "neurons," linked together like the human brain. Each neuron receives input, processes it, and passes on its output to the next layer of neurons. The strength and nature of the connections determine how the network responds to inputs, which is essentially how it learns.

What are neurons?

In the context of neural networks, a neuron is a mathematical function that collects and classifies information according to a specific architecture. The neuron's job is to take inputs, which can be the features from the data, multiply them by some weights, add them up, and then pass them through an activation function which controls the neuron's output.

How do they work together in networks?

Neurons are organized in layers that form a network. The first layer is the input layer, where the data enters the network. The last layer is the output layer, where we get our result or prediction. Between these two are hidden layers, where the actual processing is done via a system of weighted connections. The weights are adjusted during training — this is the learning in "deep learning." As data flows through the network, each layer's output becomes the subsequent layer's input, culminating in an output that provides the prediction or classification.

Layers in Deep Learning

Deep learning networks can have tens or even hundreds of layers, hence the term "deep" in deep learning. Each layer transforms its input data into a slightly more abstract and composite representation.

  • Input Layer : The input layer is where the data enters the neural network. The number of nodes in this layer corresponds to the number of input features. For instance, in image recognition, each pixel in the image data becomes an input node.
  • Hidden Layers : Hidden layers are where the complex processing happens through a system of weighted "connections." Each hidden layer's neurons take the outputs of the previous layer, apply a set of weights to them, and pass them through an activation function. The layers "learn" by adjusting these weights based on the error of the output compared to the expected result.
  • Output Layer The output layer is the final layer. The number of neurons corresponds to the number of outputs the network is designed to produce. For example, in a classification task, each neuron in the output layer represents a class, and the neuron with the highest value indicates the predicted class.

Activation Functions

Activation functions are critical to neural networks' ability to make complex decisions and learn from data. They decide whether a neuron should be activated or not, influencing the output of the model.

  • Sigmoid : The sigmoid function is a widely used activation function, historically popular for its nice analytical properties. It takes any real-valued number and maps it into a value between 0 and 1, making it useful for models where we need to predict the probability as an output. However, it's not commonly used in deep networks due to problems like vanishing gradients, where gradients get smaller and smaller as we go back through the network, slowing down learning.
  • ReLU : The Rectified Linear Unit (ReLU) has become the default activation function for many types of neural networks because it solves some problems of the sigmoid function. It's defined as f(x)=max(0,x). Essentially, if the input is positive, the output is that same number; if the input is negative, the output is zero. This simplicity leads to faster training and has been shown to give better performance for many types of networks.


There are many other activation functions, each with its own characteristics and uses. Some notable ones include:

  • Tanh (Hyperbolic Tangent): Similar to the sigmoid but maps values between -1 and 1.
  • Softmax: Often used in the output layer of a classifier, this function outputs a probability distribution over multiple classes.
  • Leaky ReLU: A variation of ReLU that allows a small, positive gradient when the unit is not active.

the basics of deep learning revolve around understanding how neural networks are structured and function. The layers of neurons, connected and weighted, learn from data by adjusting these weights, all guided by activation functions that dictate the flow of data through the network. This intricate yet elegant system forms the backbone of the most advanced AI applications today, from voice recognition systems to medical diagnosis tools. As technology and understanding of these networks improve, so too will the capabilities and applications of deep learning.

Key Concepts and Architectures in Deep Learning

Deep learning's versatility and power come from its various architectures and learning paradigms. Each has unique characteristics suited to different types of problems. Understanding these is crucial for anyone delving into this field.

Learning Paradigms

  • Supervised Learning : Supervised learning involves learning a function that maps an input to an output based on example input-output pairs. It infers a function from labeled training data consisting of a set of training examples. In supervised learning, each example is a pair consisting of an input object (typically a vector) and the desired output value (also called the supervisory signal). A supervised learning algorithm analyzes the training data and produces an inferred function, which can be used for mapping new examples.
  • Unsupervised Learning : Unsupervised learning is where you only have input data (X) and no corresponding output variables. The goal for unsupervised learning is to model the underlying structure or distribution in the data in order to learn more about the data. It's used against data that has no historical labels. The system isn't told the "right answer." The algorithm must figure out what is being shown. The goal is to explore the data and find some structure within.
  • Reinforcement Learning : Reinforcement learning is a type of machine learning algorithm that is based on the idea that agents learn how to behave in an environment by performing actions and seeing the results. In reinforcement learning, an agent makes observations and takes actions within an environment, and in return, it receives rewards. Its objective is to learn to act in a way that will maximize its expected long-term rewards.

Neural Network Architectures

  • Convolutional Neural Networks (CNNs) : CNNs are powerful for processing data that has a grid-like topology, such as images. A CNN convolves learned filters over the input data, uses pooling to downsample the data, and then applies fully connected layers to derive predictions. It's particularly known for its ability to pick out hierarchical patterns — low-level features like edges in earlier layers, and high-level features like shapes or objects in later layers.
  • Recurrent Neural Networks (RNNs) : RNNs are networks with loops in them, allowing information to persist. In an RNN, connections between nodes form a directed graph along a temporal sequence. This allows it to exhibit temporal dynamic behavior. Unlike feedforward neural networks, RNNs can use their internal state (memory) to process sequences of inputs. This makes them applicable to tasks such as unsegmented, connected handwriting recognition or speech recognition.
  • Long Short Term Memory Networks (LSTMs) : LSTMs are a special kind of RNN, capable of learning long-term dependencies. They were introduced to avoid the long-term dependency problem, remembering information for long periods as the default behavior. LSTMs have a chain-like structure, but the repeating module has a different structure. Instead of having a single neural network layer, there are four, interacting in a very special way.
  • Autoencoders : An autoencoder is a type of artificial neural network used to learn efficient codings of unlabeled data (unsupervised learning). The aim of an autoencoder is to learn a representation (encoding) for a set of data, typically for the purpose of dimensionality reduction. Traditionally, autoencoders were used for this purpose, but nowadays, they're being used for more diverse tasks, such as generative modeling.
  • Generative Adversarial Networks (GANs) : GANs are an approach to generative modeling using deep learning methods, such as CNNs. Generative modeling is an unsupervised learning task in machine learning that involves automatically discovering and learning the regularities or patterns in input data in such a way that the model can be used to generate or output new examples that could have been drawn from the original dataset. GANs consist of two models, a generator and a discriminator, that are trained simultaneously through a competitive process: the generator tries to produce data that come from some probability distribution, and the discriminator tries to determine whether the data it receives is from the model distribution or the true data distribution.

the field of deep learning encompasses a wide variety of architectures and learning paradigms, each suited to different types of data and problems. Understanding these key concepts and how they differ and can be applied is essential for anyone looking to delve into or work with deep learning. As the field continues to evolve, so too do these architectures, becoming more sophisticated and adapted to a broader range of tasks. Whether it's recognizing speech, translating languages, driving a car, or creating new, synthetic images, deep learning is at the forefront, pushing the boundaries of what machines can learn and do.

Deep Learning

Tools and Frameworks in Deep Learning

Deep learning has seen a rapid evolution, not just in terms of research and algorithms but also in the tools and frameworks developed to make these algorithms accessible and scalable. These tools are designed to facilitate the design, training, and deployment of deep learning models. Here are some of the most prominent ones:

  • TensorFlow: TensorFlow, developed by the Google Brain team, is an open-source library for numerical computation and machine learning. TensorFlow provides a comprehensive, flexible ecosystem of tools, libraries, and community resources that lets researchers push the state-of-the-art in ML, and developers easily build and deploy ML-powered applications. It supports a range of tasks but has a particular focus on training and inference of deep neural networks. TensorFlow is known for its powerful computational graph abstraction, allowing developers to define data flow graphs to structure their models. This flexibility means that you can create complex architectures with relative ease. TensorFlow also offers TensorBoard, a tool for visualization that makes it easier to understand, debug, and optimize TensorFlow programs.
  • PyTorch is an open-source machine learning library based on the Torch library, used for applications such as computer vision and natural language processing. Developed by Facebook's AI Research lab, it is known for its ease of use, dynamic computational graph, and strong support for CUDA which makes it easier to compute on GPUs. PyTorch is particularly favored for academic research and prototyping due to its intuitive and dynamic nature. The dynamic computation graph is a defining feature that allows you to change the way your network behaves on the fly and understand it with ease, unlike static graphs in TensorFlow.
  • Keras is an open-source software library that provides a Python interface for artificial neural networks. Keras acts as an interface for the TensorFlow library. Up until version 2.3, Keras supported multiple backends, including TensorFlow, Microsoft Cognitive Toolkit (CNTK), and Theano, making it more flexible for the users to switch between platforms. Keras is known for its user-friendliness, modularity, and extensibility. It allows for easy and fast prototyping and supports both convolutional networks and recurrent networks, as well as combinations of the two. Keras is particularly popular among beginners due to its easy-to-use API.
  • Scikit-Learn: While not deep learning-specific, Scikit-Learn is an important tool in the machine learning ecosystem. It's a Python library that offers simple and efficient tools for data mining and data analysis and is accessible to everybody and reusable in various contexts. It's built on NumPy, SciPy, and matplotlib and provides simple and efficient tools for predictive data analysis.
  • Apache MXNet is an open-source deep learning framework designed for efficiency and flexibility. It allows you to mix symbolic and imperative programming to maximize efficiency and productivity. In its core, MXNet contains a dynamic dependency scheduler that automatically parallelizes both symbolic and imperative operations on the fly. It's a choice for both efficiency and flexibility.
  • Fast.ai is a research group with a deep learning library that sits on top of PyTorch, and it's designed to simplify the process of obtaining state-of-the-art results in standard deep learning domains. Fast.ai's library provides high-level components that can quickly and easily provide state-of-the-art results in standard deep learning domains, and low-level components that you can mix and match to build new approaches.

the choice of a deep learning framework depends on the specific needs of the user, including ease of use, scalability, and the specific type of model they are looking to build. While TensorFlow and PyTorch are the most widely used due to their extensive features and community support, other tools like Keras offer simplicity and ease of use, making them popular among beginners. As the field of deep learning continues to grow, these tools and frameworks evolve, providing more features, simplifying complex tasks, and making deep learning more accessible to a broader audience.

Applications of Deep Learning

Deep learning has revolutionized numerous fields, providing ways to approach problems that were once considered intractable or requiring intense human effort. Here are some of the key areas where deep learning has made significant impacts:

Image and Speech Recognition

  • Image Recognition: Deep learning particularly shines in image recognition tasks, significantly outperforming other machine learning approaches. Convolutional Neural Networks (CNNs) are widely used for image classification, object detection, image segmentation, and other computer vision tasks. They can identify faces, individuals, street signs, tumors, platypuses, and many other aspects of visual data. The success of deep learning in image recognition is one of the main reasons why the world has woken up to the potential of deep learning.
  • Speech Recognition: Deep learning has significantly improved the ability of machines to understand and respond to voice commands and dictations. Recurrent Neural Networks (RNNs) and Long Short Term Memory networks (LSTMs) are particularly useful in this domain due to their ability to work with sequences of data. Today, deep learning models are at the heart of voice-activated systems like Google Assistant, Siri, and Alexa.

Natural Language Processing (NLP)

NLP involves the interaction between computers and humans using the natural language. The ultimate objective of NLP is to read, decipher, understand, and make sense of human languages in a valuable way. Deep learning has improved machine translation, language modeling, and other NLP tasks. Models like Transformers and BERT (Bidirectional Encoder Representations from Transformers) have set new standards for accuracy in tasks like sentence classification, question answering, and more, making machines better at understanding and generating human language.

  • Autonomous Vehicles: Deep learning is a key technology behind driverless cars, enabling them to recognize a stop sign, or to distinguish a pedestrian from a lamppost. It is being used for detecting objects (cars, pedestrians, etc.), predicting the behavior of other drivers and pedestrians, and planning a path to the destination. These systems rely heavily on deep learning models to detect, classify, and react to countless objects while driving on the road.

  • Healthcare: In healthcare, deep learning is doing wonders. It's being used for a variety of tasks including medical image analysis, drug discovery, and genomics. Deep learning models help in diagnosing diseases from X-rays, MRIs, and CT scans. They are also used in predictive analytics to help plan patient interventions, understand disease progression, and more. Personalized medicine is another area where deep learning algorithms are making a significant impact, offering treatment plans tailored to individual patients' genetic makeup, lifestyle, and other factors.

  • Finance: Deep learning is also transforming the finance industry. It's used for algorithmic trading, portfolio management, fraud detection, loan underwriting, and risk management among other things. In algorithmic trading, deep learning models are used to predict stock movements and make trading decisions. For fraud detection, deep learning is used to identify unusual patterns of transactions which may indicate fraudulent activity. Banks and other lenders use deep learning to assess the risk of loaning money to individuals or businesses, improving the accuracy of credit scoring models.

Deep learning has a wide array of applications, many of which are becoming integral to our daily lives. From the way we interact with our devices through voice and image recognition to significant advancements in healthcare and finance, deep learning is enabling rapid progress across sectors. As data continues to grow and computing power increases, the scale and impact of deep learning applications are only expected to increase, driving innovation and improving efficiencies across industries. The versatility and robustness of deep learning models, coupled with the continuous development in algorithms and an increase in data availability, ensure that deep learning will continue to push the boundaries of what's possible in AI.

Challenges and Future of Deep Learning

Deep learning has made significant strides in recent years, but it's not without its challenges. As we look to the future, understanding these challenges and the potential direction of deep learning is crucial.

Data Requirements

Deep learning models are notoriously data-hungry. They require massive amounts of labeled data to train on and can be significantly less accurate if the data is of poor quality or not representative of the real-world scenario. Obtaining such large datasets can be difficult and expensive, and labeling them can be time-consuming and prone to human error. Furthermore, in domains like healthcare or finance, data might be highly sensitive, adding layers of complexity regarding privacy and security.

Future Trends and Predictions:

  • Semi-supervised and Unsupervised Learning: Advances in these areas could reduce the dependency on large labeled datasets.
  • Data Augmentation and Synthetic Data: Techniques to artificially expand the dataset might become more sophisticated, making training deep learning models more feasible with less data.

Computing Power

Deep learning models, especially the state-of-the-art ones, require significant computational power. Training large models can take days or even weeks, even on powerful GPUs or TPUs. This not only slows down the research and development cycle but also makes it expensive and less accessible to individuals or smaller organizations.

Future Trends and Predictions:

  • Efficient Model Architectures: There's ongoing research into creating more efficient models that require less computational power without compromising performance.
  • Distributed and Collaborative Learning: Methods like federated learning allow for distributed model training across multiple devices, potentially reducing the need for centralized, powerful computing resources.

Ethical Considerations

As deep learning systems become more prevalent, ethical considerations are increasingly important. Issues include bias in AI, where models might perform differently for different demographic groups, often as a result of biased training data. There's also the concern of job displacement as AI systems become capable of performing tasks traditionally done by humans. Moreover, the use of deep learning in surveillance and other sensitive areas raises privacy concerns.

Future Trends and Predictions:

  • Ethical AI Frameworks: More robust frameworks and guidelines for ethical AI development are likely to be established.
  • Explainable AI: Efforts to make AI decisions more transparent and understandable to humans will be crucial in building trust and managing ethical considerations.

Future Trends and Predictions

Beyond addressing these challenges, several trends are likely to shape the future of deep learning:

  • Integration with Other Fields: Deep learning will continue to intersect with other disciplines, leading to new breakthroughs. For instance, combining deep learning with quantum computing could lead to even more powerful AI systems.
  • Personalization: As deep learning becomes more sophisticated, personalized AI that understands and predicts individual preferences could become more prevalent in applications ranging from healthcare to retail.
  • Autonomous Systems: Advances in deep learning will continue to drive the development of autonomous systems, not just in vehicles but in drones, robots, and other areas.
  • AI for Social Good: There's increasing interest in using AI to address social and environmental challenges, from climate change to healthcare.

Deep learning is a field with tremendous potential but also significant challenges. As it continues to evolve, it will likely permeate even more aspects of our lives, driving innovation and perhaps even changing the way we interact with technology and each other. Addressing the challenges of data requirements, computing power, and ethical considerations will be crucial in realizing the full potential of deep learning. At the same time, staying abreast of future trends and predictions will help guide the development of this exciting field. The journey of deep learning is far from over, and its future looks as promising as it is complex.

Case Studies: Deep Learning in Action

Deep learning has been applied successfully across various sectors, demonstrating its versatility and power. Here are a few case studies that highlight its real-world impact.

1. Healthcare: Diagnosing Diabetic Retinopathy

Diabetic retinopathy is a condition that can lead to blindness if not detected early. Traditionally, detecting it requires a skilled clinician to examine and interpret digital color fundus photographs of the retina.


Researchers used deep learning models to automatically detect diabetic retinopathy in retinal images. The deep learning system was trained with a large set of fundus images graded for disease presence and severity. The model learned to identify signs of diabetic retinopathy with accuracy comparable to human experts.


The deep learning system demonstrated the potential to automate the screening process for diabetic retinopathy, making it more efficient and accessible. This is particularly beneficial in areas where skilled clinicians are scarce.

2. Autonomous Vehicles: Waymo's Self-Driving Cars

Autonomous vehicles have the potential to reduce accidents, increase transportation efficiency, and revolutionize mobility. Achieving this requires the vehicle to understand and navigate complex environments.


Waymo, a project within Alphabet, uses deep learning for its self-driving vehicles. The technology helps cars understand their surroundings, predict what others will do, and decide on a safe and efficient course of action. This includes detecting pedestrians, reading traffic lights, and navigating urban streets or highways.


Waymo's vehicles have driven millions of miles on public roads, continually learning and improving. While fully autonomous vehicles are still a work in progress, Waymo's successes demonstrate the potential of deep learning in making self-driving cars a reality.

3. Finance: Fraud Detection at PayPal

Fraudulent transactions are a significant issue for financial institutions. Traditional methods of detection involve setting hard rules, but these can be circumvented and often result in high false-positive rates.


PayPal uses deep learning to identify potentially fraudulent transactions. The system analyzes vast amounts of transaction data and learns to identify patterns indicative of fraud. It can adapt to new types of fraud, reducing false positives and catching fraudulent activity more efficiently.


PayPal has reported improved accuracy in detecting fraudulent transactions, leading to increased trust and safety for its users. The system's ability to adapt to new fraud patterns makes it a robust solution in the ever-evolving landscape of online transactions.

4. Retail: Personalized Recommendations at Amazon

Personalized recommendations can significantly enhance customer experience and drive sales. However, understanding individual preferences and behaviors is a complex task.


Amazon uses deep learning to power its recommendation engine, analyzing customer behavior, browsing history, and purchase history. The system identifies patterns and predicts what products individual customers are likely to be interested in.


The personalized recommendation system has been a key factor in Amazon's success, driving sales and customer satisfaction. It demonstrates how deep learning can be used to understand and predict human behavior at an individual level.

These case studies are just a few examples of how deep learning is being applied in the real world, driving innovation and solving complex problems. They demonstrate not only the versatility of deep learning across different sectors but also the tangible benefits it can bring. As the technology continues to evolve, it's likely that deep learning will play an even more significant role in shaping our world. Each success story also brings valuable lessons in the importance of quality data, the need for robust and adaptable models, and the potential ethical considerations as we integrate deep learning more deeply into our lives.

Conclusion: Embracing the Future with Deep Learning

Deep learning stands as one of the most transformative technologies of our time, driving changes across industries and reshaping our understanding of what machines can do. As we've explored, its applications range from diagnosing diseases and driving cars to enhancing financial security and personalizing customer experiences. The journey of deep learning is marked by both its profound impact on practical applications and the continuous exploration and expansion of its capabilities.

Recap of Deep Learning Importance

Deep learning's importance lies in its ability to learn from and make sense of vast amounts of data, uncovering patterns and insights that are often infeasible for humans to discern. Its strength in handling complex, high-dimensional data across various domains—from images and text to sound and sequences—makes it a versatile tool for tackling a wide array of problems. The success stories in healthcare, autonomous vehicles, finance, and more are testaments to its transformative power. However, with great power comes great responsibility. The ethical considerations, data requirements, and need for computational resources highlight the challenges accompanying deep learning's widespread adoption.

Encouragement for Continued Learning and Exploration

The field of deep learning is continually evolving, with new models, techniques, and applications emerging regularly. For those involved or interested in this field, the journey is as exciting as it is endless. Here are a few encouragements for continued learning and exploration:

  • Stay Curious: The landscape of deep learning is vast and ever-changing. Stay curious and open to new ideas, techniques, and domains where deep learning can be applied. Continuous learning is key in a field that evolves as rapidly as AI and deep learning.
  • Engage with the Community: The deep learning community is robust, inclusive, and continually growing. Engage with forums, attend conferences, or contribute to open-source projects. Collaboration and knowledge sharing are pivotal for innovation and understanding.
  • Ethical and Responsible AI: As you delve deeper into deep learning, consider the ethical implications of your work. Strive to develop and apply deep learning responsibly, ensuring that the benefits are widespread and that the technology is inclusive and fair.
  • Hands-On Practice: Theory is crucial, but so is practice. Work on projects, participate in competitions like those on Kaggle, or contribute to research. Hands-on experience is invaluable in understanding and leveraging the power of deep learning.

Looking Ahead

As we look to the future, deep learning is set to continue its trajectory of growth and influence. The potential for positive impact is enormous, from addressing climate change and enhancing healthcare to revolutionizing industries and creating new forms of art and communication. However, the path forward involves not just technological advancements but also thoughtful consideration of the societal impacts, ensuring that the benefits of deep learning are accessible and equitable.

In conclusion, deep learning is not just a field of study or a set of technologies; it's a catalyst for innovation and change. Whether you're a researcher, developer, business leader, or simply an enthusiast, your journey with deep learning can contribute to this wave of transformation. Embrace the challenges, celebrate the successes, and continue to learn and explore. The future is not just shaped by what deep learning can do, but by what we, as a global community, choose to do with it.

Also, read - Dive into Apache Parquet: The Efficient File Format for Big Data

Ashish Kasama

Co-founder & Your Technology Partner

Ashish KasamaAshish KasamaAshish KasamaAshish Kasama

One-stop solution for next-gen tech.

Related Blogs

The latest from our innovation team


Uncover Hidden

Growth Opportunities

With Data Science, AI & ML.

Get a FREE Consultationarrow


We offer our expertise in a number of business domains. We try to deliver the best services right from pre built platforms to custom solutions.

10+ years of industry experience

1000+ global base of customers

500+ qualified resources

Transparent cost

Get a quote now !


100% confidential and secure

Read *T&C

Our Global Footprint

Lucent innovation
We are a family filled with talented experts that help global brands, enterprises, mid-size businesses or even startups with innovative solutions.
Accelerate your tech projects with us. Get in touch with us.
Follow us
lucent innnovation: cltuch

Lucent Innovation, © 2024. All rights reserved.Privacy policyDMCA compliant image