Project Information

Introduction to Intelligent Systems (430.457) Fall 2014

Pedestrian Detection Contest (12/5 - 12/12)


  • Submit your project at any time between 12/5 and 12/12 to improve your ranking on the list. Submit via eTL or send a ZIP file to TAs.
  • The ranking with scores will be updated daily. 
  • The final ranking after 12/12 will be used for grading. 
  • When you submit your new code, if the score of your new code is worse than the previous submission, the previous score will be posted on the ranking. In addition, you will be notified about the new score, so you can make a proper adjustment. Hence, we encourage you to submit a newer version whenever possible.

Final Report (due: 12/19 23:50)


  • Write a final report per team explaining your codes
    • i.e. What you did to improve the performance, Which algorithm/heuristic methods were successful while doing this term project
  • Final port will be also considered in your final score.
  • There is no fixed format or number of pages to fill in, write as freely as you can
  • Upload your report on ETL, until 12/19 23:50 KST

Final Ranking (last updated 12/13 21:53 KST)


  • Team F has made a new record!
  • Grade_20141213_2
  • We finally finished to grade all submitted codes! 
  • Only maximum value of F1-score has been updated to above list! 
    • Ask TA via E-mail ( hyemin.ahn@cpslab.snu.ac.kr ), if you want to know about the your detector's f1-score.

Final Assignment (second due: 12/4)


Submit the improved version of your pedestrian detector.

  • Never use the extract_hog.p file that we've provided in Assignment 2. We'll give a penalty to a team who didn't use their own feature extractor.

Results

  • Final_grade_2
  • Team J, O, Q    : Not submitted.
  • Team A : Delayed

 

Final Assignment (first due: 11/27)


In this assignment, you need to build a pedestrian detector completely.

Results 

Final_grade_1

  • Team N                 : Delayed for 2 hours.
  • Team J, M, O, Q    : Not submitted.
  • Team H                 : Used extract_hog.p file that we've distributed. 
  • TA's Score             : F1 score : 0.6838 / Precision : 0.6349 / Recall : 0.7407

Assignment 2 (due: 11/20)


In this assignment, you need to understand how to detect people in given test images with the given trained SVM model.

Assignment 1 (due: 11/13)


The objective of this semester’s term project is to build a detector, which can find pedestrians from an image using a supervised learning technique. This assignment 1 is a preparation for this term project. In this assignment, you need to understand HOG and implement codes for extracting a HOG feature from an image patch.

Teams


Team Member 1  Member 2 
A Ilse Mireya Alejo Guitierrez Beatriz Trinidad Santos Buitrago 
B Seungwon Lee  Kyuhong Sim 
C Minsu Kwon  Jaeyoon Yoo 
D Gyumin Oh  Jungkwon Lee 
E Jinhwan Park  Junwoo Lee 
F Jeongwoo Kim  Juyong Kim 
G Jihoon Kim  Jungyeol Kwon 
H Dongyoon Kim Sungwook Choi
I Igor Ferreira Pinto  Vincent Perot
J Felipe Ergueta Luana Constantivo
K Hanseul Lee Doosan Baek 
 L Seunggyu Chang -
M
Jangseop Shin -
 N Hyunki Hong  -
 O Guerin Erwan  -
 P  Donghoon Lee  -
 Q  Zhang Guangxue  -