The problem Fashion recommendation system using deep learning solves

In context of Nepal, There are very few e-commerce websites/app which provide authenti recommendation to its user. Also, The recommendation system which they have is too old and they are not
matched with the current trend . Suppose, There is a person whose name is ram . He went to his friends house and say a cup . He also want similar cup. But he dont have any clue about the
cup also the cup dont have any label or brand which can tell us about the cup. By the help of our website if he upload the image of this cup then he may get the exact cup or similar type of product
So our website and model is user friendly and it even saves the time of the user. Searching the method of product by typing keywords have become a traditional way . In this modern era,
By the help of AI we can simply create a change in the ecommerce industry . In Nepal there are few/no any website which provide this feature of searching the product by uploading the image.
So, We have come up with the model where user/customer can upload there desirable fashion items in our website and then our model will show them best 5 similar recommednations .

Challenges we ran into

We had numerous difficulities throughout the hackathon, which we overcame by spending countless hours googling and debugging . One of the main issue we ran into was - The cv2.imshow() and cv.imshow() functions from the opencv-python package were incompatible with Jupyter notebook. So we have come up with the alternative to use module -
import from google.colab.patches import cv2_imshow and display using cv2_imshow(<image>) .
However , we have also used python library called matplotlib to resize , read and display image.

As we know that we cannot host our website on cloud due which we need export all the dataset and trained model
files i.e embedding ,file name pickly files in our local machine. While doing that we encountered a tensorflow version mismatch error. How ever we fixed this issue taking help of the mentors and
stack overflow. After that we have done backend in the django frame work . There was a slight bug regrading the
media root directory path in django.