ISSM AI Solution Contests
ISSM SEM Image Classification AI Algorithm Contest
This contest aims to broaden the scope of practical research and development through the trial using actual data generated at the semiconductor manufacturing. This contest competes the algorithm for "defect/particle classification in SEM images" that is essential for improving the yield of semiconductor manufacturing. Approximately 4,00 particle SEM images occurred in actual semiconductor manufacturing will be provided through the cooperation by ISSM committee members. The contest participants are required to create a learning model that automatically classifies about the defect/particle SEM images. The particle area identification and classification accuracy by the applicants will be reviewed comparing with the classified list made by professional engineers in semiconductor manufacturing.
Entry
Entry has been closed
Requirement
- Qualification: Students (Both individuals and teams can participate)
- Kaggle in Class: The participants are required to use Kaggle in Class, a tool used by hundreds of universities around the world to practice data analysis techniques. It is encouraged for those who want to try Kaggle-in-class for the first time.
- Language: Python
- Criteria: The SEM images of defects and particles are classified into the designated classes. The ranking is competed by the classification accuracy rate of the answers.
- Required skills: In addition to the methods of basic image recognition/classification , it is necessary to deal with data/issues that occur during actual operation at semiconductor manufacturing fabs, such as data imbalance, unclear classes, and micro defects.
- Submission: The applicants are required to submit the following report and abstract for preliminary experiments.
1. Report on algorithm code to be submitted to Kaggle-in-Class
2. Abstract (1page of MS Word, template) to be submitted to ISSM secretariat - Data sets: SEM images actually used for semiconductor manufacturing are provided.
1. Learning data for preliminary experiments: 3,400-4,000 samples
2. Data for final selection: 200-300 samples*
(*It may contain the samples misclassified in preliminary experiments.) - Note: The provided data sets are ONLY allowed to be used for ISSM SEM Image Classification AI Algorithm Contest.
(ISSM濮斿摗浼佹キ銈堛倞鎻愪緵)
Submission
Upload your abstract (word file in ISSM template) to the following Drop Box.
The file Name must have your team name.
eg. Team-ISSM. docx
ISSM SEM Image Classification AI Algorithm Contest
Schedule (The deadline has been extended)
- Entry due date: November 13, 2020
*It is highly recommended to complete entry sheet so that you can get the data with Kaggle-in-class ID. - Kaggle User application date: From September 15, 2020
- Report on Algorithm code due date: December 4, 2020
- Abstract due date: December 4, 2020
Award and Award Examination Guidelines
Award: The Best Excellent Award and The Technical Awards will be recognized at the Award Ceremony during ISSM 2020 conference period (December 15-16, 2020).