Click an Image and Display Similar Results

General Discussion
,
#1

Hello Everyone :slight_smile:

Hi, so i have been building a Django project in python for the last two months and have got everything ready from user registration to file upload. Now my LAST part left is the following.

A User clicks an image on site which then drops down and shows similar results of that image.

How can i achieve that ? I want to fetch the results from the Database back-end of python (Will use the async Fetch function for that ) and the add Event listener for clicking an Image.

Will Really Need Your Help. Thanks :slight_smile: Will TensorFlow.js be suitable? I am a newbie at this.

So far I know the Following:

i have got the concept, that is extract the features and then compare histograms and display similar results.

is there an example in java script. i tried finding it. just could not get it any where :frowning:

mostly i found is image resizing

Will it be better to do it in the back end or use JavaScript ? Sorry for Noob Question

1 Like
#2

You’re in the right direction regarding the features.

My first try would be to extract the image feature vector and use it (instead of the text embeddings) shown in this tutorial: Semantic Search with Approximate Nearest Neighbors and Text Embeddings  |  TensorFlow Hub

4 Likes
#3

hi thanks you so much for the reply. I will check that link out,

Hey do you have any experience with opencv?

SO far i have got the edges of the two images using canny method, but now what next should i do to compare my images ?

Here is the code, Can you Please Help Out:

import cv2 
import os
 
import numpy as np
 
import matplotlib.pyplot as plt
 
img1 = cv2.imread("1-Flip-flops.jpg",0)
img2 = cv2.imread("29-Leather mules.jpg",0)
 
edges1 = cv2.Canny(img1, threshold1=30, threshold2=100)
edges2 = cv2.Canny(img2, threshold1=30, threshold2=100)
 
print(edges1)
 
plt.imshow(edges1, cmap = "gray")
plt.show()
 
plt.imshow(edges2, cmap = "gray")
plt.show()
1 Like
#4

Try to give pass to

It is not clear what is your specific context so this tutorial is about image retrieval with landmarks and it could be a good baseline.

See also on the same TFHub model:

3 Likes
#5

Hi thanks for the recommended article, the nearest neighbor i have to find for similar images.

Can you also check my opencv code above ? I am trying to match or find similar images of shoeswear that I scrapped from Shoe retailers online. I have detected the edges but now what?

Sorry if wrong forum,

1 Like
#6

You need to follow and understand some basic examples e.g. with MNIST and then reproduce it for your dataset.

For a complete basic example see:

2 Likes