You are on page 1of 3

 

Hiring challenge 
 

Question: 
Image  quality  detection  has  always  been  an  arduous  problem  to  solve  in 
computer vision.  
 
In  this  assignment  you  have  to  come  up  with  features  and  models  to  build 
a classifier which will predict whether a given image is blurred. 
 
Heads-up:  You  can  train  CNN  if  you  want,  but  beware  of  the  time  running 
out. 

Dataset: 
http://mklab.iti.gr/files/imageblur/CERTH_ImageBlurDataset.zip 
Note:  
1) Details about the dataset is present in README.txt inside the zip. 
2) The dataset is of ~3.7 GB.  
 
 

 
  
 

Instructions 
1. To  submit  your  model,  train  script,  test  script  and  support  files,  Zip  it 
into  the  folder  and  name  the  zip  folder  with  your  Name  followed  by 
accuracy  on  evaluation  set  (eg.  Shashank_gupta_89.zip).  [  you  can 
round off your accuracy while naming the zip folder ] 
2. Share  your  zip  file  over  the  google  drive  with  access  to 
shashank.gupta@cloudsek.com​ and ​bofin.babu@cloudsek.com  
3. The  test  script  should  log  the  accuracy  of  your  model  on  the 
evaluation set. 
4. Your  submission  should  contain  a  detailed  README.md,  to  help  us 
understand and run at our end.  
5. Don’t  include  any  library  files,  instead  provide  instructions on how we 
can install at our end. 

Judgement: 
Both  Machine  Learning  and  Computer  Vision/Image  Processing  skills  will 
be judged 
 
1. Novelty of thinking involved in feature engineering. 
2. Accuracy of the classification model. 
3. Inference time 

Constraints: 
You  are  free  to  use  your  favourite  libraries/frameworks.  Kindly  avoid  using 
Matlab or any paid tools. 
  
 

Submission: 
Within 48 Hours from when the assignment was received. 
 

References: 
You may refer to this tutorial 
https://www.pyimagesearch.com/2015/09/07/blur-detection-with-opencv/ 
(Don't copy the same) 

You might also like