LabAssignment1Report - aaz000966/CS5542 GitHub Wiki

Lab Assignment #1 Report Zakari, Abdulmuhaymin (29)

This is a lab assignment report for performing NLP and sift techniques on a dataset

• NLP: I used NLTK, an open source python NLP library that has great potentials. I tried to use spacy since it’s provisioned as much faster library, but I faced several issues, the following code contains comment on the workflow.

  1. #This code was inspired direclty by the ICPs given in the class
  2. #i'll be using
  3. import nltk
  4. from nltk.corpus import stopwords
  5. from nltk.stem import WordNetLemmatizer
  6. #I tried to use spacy, but I faced some issues
  7. import spacy
  8. from nltk.tokenize import word_tokenize
  10. #to count how many words and lines in the file, helped during the development
  11. Words=0
  12. Lines=0
  14. #linking the dataset from the project folder and read the file
  15. file= open("SBU_captioned_photo_dataset_captions.txt","r")
  17. data=file.read()
  19. #simple iteration to count the content
  20. with file as f1:
  21. for line in f1:

  22.     Lines += 1

  23.     for word in line.split():

  24.         Words += 1
  26. print(Words)
  27. print(Lines)
  29. #to creat a newfile with the output required
  30. TokenF=open("Tokenized.txt", "w")
  32. #tokenizing the edetatset using word_tokenize(str), a predefined nltk method
  33. TokenF.write(str(word_tokenize(data)))
  36. #creating a new output file for lemma operation
  37. LemmaF=open("Lemmatized.txt", "w")
  38. lemma = WordNetLemmatizer()
  41. file= open("SBU_captioned_photo_dataset_captions.txt","r")
  42. #for the sake of demonstration, I'm only writng the words with different architecture
  43. with file as f2:
  44.  for line in f2:

  45.      for word in line.split():

  46.          if word != lemma.lemmatize(word):

  47.              LemmaF.write(str(word + " : " + lemma.lemmatize(word) + "\n"))
  50. print("done")

As it’s shown, the first part of this program produces a text file with the dataset segmented or tokenized:

The second part produces a file of the lemmatized version of the dataset:

• Sift:

The other part of the assignment is to sample some images and process them through sift. I used 2 images that reflect my project theme (food) and take benefits of the sift image processing approach. The code as follow:

  1. #This code was inspired direclty by the ICPs given in the class
  5. #Importing cv2 ,matplotlib
  6. import cv2
  7. import matplotlib.pyplot as MyPyplt
  12. #Inserting photos to the system, sources: https://sakatavegetables.com/specialty-tomatoes/
  13. #as well as converting them to CV2 RBG
  15. MyImage1 = cv2.imread('1.jpg')
  17. MyImage1 = cv2.cvtColor(MyImage1, cv2.COLOR_BGR2RGB)
  19. MyImage2 = cv2.imread('2.jpg')
  21. MyImage2 = cv2.cvtColor(MyImage2, cv2.COLOR_BGR2RGB)
  23. #As descriped that this line is required for version-3
  24. orb = cv2.ORB_create()
  26. #detecting and computing keypoints on both images, des1,2 will be used later on.
  27. ImgKP, des1 = orb.detectAndCompute(MyImage1, None)
  29. ImgKP2, des2 = orb.detectAndCompute(MyImage2, None)
  31. #new 2 images with the keypoints displayed
  32. Img1AfrPrs = cv2.drawKeypoints(MyImage1,ImgKP,None) # Draw circles.
  34. Img2AfrPrs = cv2.drawKeypoints(MyImage2,ImgKP2,None) # Draw circles.
  36. MyPyplt.imshow(Img1AfrPrs)
  37. MyPyplt.show()
  38. MyPyplt.imshow(Img2AfrPrs)
  39. MyPyplt.show()
  42. bf = cv2.BFMatcher(cv2.NORM_HAMMING, crossCheck=True)
  44. matches = bf.match(des1, des2)
  46. matches = sorted(matches, key = lambda x:x.distance)
  48. #creating a new image with matches on both and connected using drawMatches()
  49. img_matches = cv2.drawMatches(MyImage1, ImgKP, MyImage2, ImgKP2, matches[:50], MyImage2, flags=2)
  51. MyPyplt.imshow(img_matches);
  52. MyPyplt.show()

Output:

Thank you.