Page 177 - Emerging Trends and Innovations in Web-Based Applications and Technologies

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
This innovative system optimizes Quick Mart’s marketplace
by improving product visibility, streamlining pricing, and
ensuring users have the information they need to make
confident transactions. By leveraging AI and machine
learning, Quick Mart redefines the second-hand shopping
experience.
IV. PROPOSED RESEARCH MODEL: Smart
Marketplaces for Used Goods
The proposed research leverages a convolutional neural
network (CNN) model to enhance Quick Mart’s innovative
approach to revolutionizing the second-hand marketplace.
CNNs are powerful deep learning structures extensively used
in classification tasks such as image recognition and object
detection. In the context of Quick Mart, the CNN will be

Fig. 2: Sample Images of Second-Hand Products in the employed to automate the categorization of various second-
Dataset hand products, streamlining the process for users and
enabling a seamless experience for both buyers and sellers.
Data Preprocessing
Effective data preprocessing is critical to the success of A critical aspect of Quick Mart’s platform is efficient and
machine learning models. During this phase, missing values accurate product categorization. The CNN model addresses
and redundant data are handled, and data augmentation is this challenge by classifying second-hand products into
applied to expand the dataset. Key steps include: distinct categories, such as electronics, furniture, fashion,
 Loading the Data: The dataset is loaded and split into and books, based on product images. This automatic
training and testing sets. categorization system simplifies navigation through the
platform, enhances user experience, and helps users find the
 Shuffling and Splitting: The data is shuffled and split products they need with minimal effort.
into training, validation, and testing subsets in an 80:20
ratio. CNN Model Architecture
The architecture of the proposed CNN model is designed to
 Label Encoding: Text-based labels are converted into process input images and produce output in the form of class
numerical representations using LabelEncoder. probabilities. The CNN consists of several sequential layers,
 Categorical Conversion: Labels are further converted each responsible for transforming the data and learning
into categorical format to improve model training different features of the images.
performance. 1. Conv2D Layer: The first layer of the CNN is a Conv2D
Feature Extraction layer that performs convolution on the input images
Feature extraction is a crucial step in improving model using a set of learnable filters. For this implementation,
accuracy. For Quick Mart, this process involves identifying 32 filters of size 3x3 are used, with a rectified linear unit
and isolating meaningful patterns or characteristics from (ReLU) activation function, which is widely adopted for
product images. These extracted features are key to accurate its effectiveness in promoting efficient learning.
product categorization and price prediction. Methods used 2. MaxPooling2D Layer: After the convolution operation,
for feature extraction include: a MaxPooling2D layer is employed to perform down-
 Texture Analysis: Features such as entropy and sampling. This layer selects the maximum value from a
homogeneity are analyzed to assess product condition window of size 2x2 in the input, reducing the spatial
and detect any defects. dimensions of the output and helping the model focus on
 Shape Analysis: Shape-based features help determine the most relevant features.
the quality and authenticity of the product. 3. Repeated Layers: The Conv2D and MaxPooling2D
 Histograms of Intensity: Color and pixel intensity layers are repeated with an increased number of filters
levels provide insights into the product’s condition. (e.g., 64 filters), maintaining the same kernel size and
activation function. This repetition allows the model to
 Spatial Filters: Applied to enhance image clarity and learn increasingly complex features at different levels of
highlight key product details. abstraction.

 Wavelet Transforms: These features allow for multi- 4. Flatten Layer: After the convolutional layers, a flatten
scale analysis of images, capturing fine-grained details. layer is used to transform the multi-dimensional output
into a one-dimensional array. This transformation
Classification
Product classification is performed using a Convolutional makes the data compatible with the fully connected
Neural Network (CNN), a deep learning model that excels in layers and prepares it for classification.
handling image datasets. The CNN is trained to classify 5. Dense Layer: The dense layer is fully connected and
second-hand products into appropriate categories and utilizes the ReLU activation function. This layer
predict their price ranges based on features such as processes the features extracted from the convolutional
condition, type, and brand. This classification system aids layers, combining them to make predictions regarding
sellers in setting competitive prices and assists buyers in product categories.
making well-informed purchasing decisions based on the
product's condition and description. 6. Final Softmax Layer: The final layer of the model is the
Softmax layer, which outputs class probabilities for each

IJTSRD | Special Issue on Emerging Trends and Innovations in Web-Based Applications and Technologies Page 167