3rd IAPR Summer School on Document Analysis

Theme: Deep Learning Applications for Document Analysis

Dates: 19th to 23rd August, 2019
Venue: Islamabad, Pakistan

Registration
Early Bird
(till June 15, 2019):
Student: PKR 3,500/- (USD 25/-)
Professional: PKR 15,000/- (USD 110/-)
Regular Registration
(till July 31, 2019):
Student: PKR 5,000/- (USD 35/-)
Professional: PKR 20,000/- (USD 145/-)
Late Registration
(till August 17, 2019):
Student: PKR 7,500/- (USD 55/-)
Professional: PKR 25,000/- (USD 180/-)
Summer School has been commenced and ended successfully, thank you for participating in the event!

On behalf of all the organizing team, PPRS would like to thank all IAPR, TC-10/11 management, speakers, participants and NUST-SEECS for their support in making 3rd IAPR Summer School on Document Analysis (held from August 19-23, 2019) a great success. The school was attended by over 100 participants from all over Pakistan and abroad.

Participant Institutes and Organizations

PPRS would like to extend special thanks to our esteemed speakers from around the globe to first agreeing to come to Pakistan, to deliver high quality content, and to engage in fruitful individual discussions with the school participants. School hosted a programming competition on hand-drawn sketches and a poster session. Winners of these competitions were awarded with appreciation certificates during the closing session of the school. Speakers and participants also visited nearby museums of national heritage and attended a colourful social event and a school dinner. Overall, the school provided valuable intellectual experience to all participants and will be instrumental in attracting more local researchers to the field of Document Analysis.

About This Event

We are pleased to announce that the 3​rd​ IAPR TC10/TC11 Summer School on Document Analysis and Recognition, endorsed by IAPR TC-10 (Technical Committee on Graphics Recognition) and IAPR TC-11 (Reading Systems) in association with Pakistan Pattern Recognition Society (PPRS) will be held at the National University of Science and Technology (NUST), Pakistan from Aug 19th to 23rd, 2019.


The 3​rd IAPR Summer School on Document Analysis is a premier gathering for researchers, scientists and practitioners in the deep learning community. Objective of the school is to provide the participants with different aspects and in-depth analysis of the latest research being carried out in the field of document understanding. Specialists from Germany, France, China and Pakistan will deliver high-quality content and will impart hands-on training to the school's audience at NUST, which is housed in the scenic landscapes of the second most beautiful capital of the world - Islamabad.School will provide a dazzling opportunity to participants to polish their learning skills and to develop their cognitive thinking. The speakers are recognized experts in their field and will shed light on Layout Analysis, OCR, Text Identification, Graph-based Methods and Image Interpretation and will also discuss the architectural aspects and state-of-the-art research in the field of document understanding and analysis. By the end of the school participants would be able to learn about:

  • Graph based methods
  • Connectionist Document Classification
  • Multi-dimensional Neural Recognition at industrial scale
  • OCR of Complex Scripts
  • Document Understanding

Highlights

  • Talks on Multi-dimensional Neural Recognition at industrial scale, Long Short-Term Memory Networks for Handwriting Recognition, Graph Based Methods, OCR, Document Image Processing and many other topics
  • Complimentary accommodation for international speakers and students
  • Return air fares for international speakers will be sponsored
  • Specialist in their respective fields will provide tutorials and demos
  • Programming competition

Focus Areas

The school will highlight the following core areas:

  • Document Understanding
  • Layout Analysis
  • Information Retrieval
  • Optical Character Recognition

The school is organized with the motivation to promulgate advanced machine learning and computer vision techniques and their applications. The school will be conducted by well-renowned experts and researchers in their domains.

Topics To Be Covered

The school will address the following topics:

  1. Deep Neural Networks for Document analysis: Multidimensional Neural Recognition at Industrial Scale, Deep Learning for Historical Document Analysis and Convolutional Neural Networks for Document Classification
  2. Handwriting Recognition: On-Line Handwriting Recognition, Long Short-Term Memory Networks for Handwriting Recognition
  3. Optical Character Recognition: Optical Character Recognition of Complex Scripts
  4. Camera Based Document Analysis: Camera Based Analysis of Texts and Documents
  5. Document Analysis Workflow: Opportunity Areas in Document Analysis; Present Past and Future of Document Analysis

Program

Click here to download program brochure

19th Aug (Mon) 20th Aug (Tue) 21th Aug (Wed) 22th Aug (Thu) 23th Aug (Fri)
Time Program -- SSDA2019
09:30 Registration + Arrival of participants and guests
10:00 Summer School Opening Session
10:30
11:00 Coffee Break + Group Photo  
11:30 Intro to Document Analysis (Prof. Dr. Cheng-Lin Liu)
12:00
12:30
13:00 Lunch
13:30
14:00 Intro to Document Analysis (Prof. Dr. Cheng-Lin Liu)
14:30
15:00
15:30 Coffee Break
16:00 Competition Overview (Prof. Dr. Imran Siddiqui)
16:30
17:00 Programming Competition (Prof. Dr. Imran Siddiqui)
17:30
18:00
18:30
Time Program -- SSDA2019
09:30 Intro to ML and Neural Networks (Prof. Dr. Marcus Liwicki)
10:00
10:30
11:00 Coffee Break
11:30 Intro to CNN (Prof. Dr. Adrian Ulges)
12:00
12:30
13:00 Lunch
13:30
14:00 Applications of CNNs (Prof. Dr. Ulrich Schwanecke)
14:30
15:00 AMIGO - Automatic Indexing of Lecture Footage (Prof. Dr. Adrian Ulges)
15:30 Coffee Break
16:00 Programming Competition (Prof. Dr. Imran Siddiqui)
16:30
17:00
17:30
18:00
18:30
Time Program -- SSDA2019
09:30 Optional Toursim + Recreational Day
10:00
10:30
11:00
11:30
12:00
12:30
13:00 Lunch
13:30
14:00 Workshop: Water resource estimation by Satellite-Based Image analysis

  1. Prof. Dr. Adrian Ulges
  2. Prof. Dr. Ulrich Schwanecke
  3. Prof. Dr. Marcus Liwicki
  4. Prof. Dr. Faisal Shafait
  5. Assis. Prof. Dr. Muhammad Shahzad
  6. Assis. Prof. Dr. Muhammad Imran Malik
14:30
15:00
15:30
16:00
16:30
17:00
17:30
18:00
18:30
Time Program -- SSDA2019
09:30 LSTM – The Basics and Current Trends (Prof. Dr. Marcus Liwciki)
10:00
10:30
11:00 Coffee Break
11:30 LSTM for Handwriting Recognition (Prof. Dr. Marcus Liwciki)
12:00
12:30
13:00 Lunch
13:30
14:00 Document Classification and Understanding (Dr. Sheraz Ahmed)
14:30
15:00
15:30 Coffee Break
16:00 Poster Session (Dr. Faisal Rashid)
16:30
17:00 Programming Competition (Prof. Dr. Imran Siddiqui)
17:30
18:00
18:30
Time Program -- SSDA2019
09:30 Open Research Directions of Document Analysis and Recognition (Prof. Dr. Seiichi Uchida)
10:00
10:30
11:00 Coffee Break
11:30 Document Analysis Research in Pakistan (Local Organizers)
12:00 Students' Presentation
12:30
13:00 Lunch
13:30
14:00 Panel Discussion
14:30
15:00 Programming Competition Results
15:30 Coffee Break
16:00 Summer School Closing Session
16:30
17:00
17:30
18:00
18:30

Poster Competition

SSDA 2019 provides students and researchers to present their concluded or ongoing research in the domain of document analysis. Any research done that falls under the scope of the summer school can be presented in this session. This is also an amazing opportunity for young researchers to present their (future) undergraduate or graduate thesis ideas and receive valuable feedback.

Instructions for Authors

Guidelines for Abstract

  • Extended abstracts should provide the detailed overview of already concluded or ongoing research.
  • Abstracts must be submitted in either Word doc. or PDF file formats. Times New Roman, no smaller than 12 pt. font, single spacing.
  • Following information should be provided in a clear and conscise way. Title of the Abstract, Author name(s), Affiliation(s), References.

Content of the Abstract

A few things that can be done for better abstract:

  • Make a case for your work by stating the problem or research question.
  • State the main point or argument of the proposed presentation.
  • State the relevance of your idea to society, past work. State the impact that your work can make.
  • If your research presents the results, provide them along with the nature and design of the experimental setup. Explain the methodology used for concluded research.

Guidelines for Camera Ready Poster

  • Poster should be submitted in A1 size(Landscape/portrait) with margins no smaller than 1/2 inch and font no smaller than 16 point.
  • Design of the poster is upto the author but it should be readable and concise. (For reference see the sample provided).
  • Header of the poster should contain title, authorship, affiliations and SSDA19 logo.
  • SSDA 2019 will print all the posters for free.

Important Dates

  • Proposal submission deadline: August 9, 2019
  • Notification of acceptance: August 12, 2019
  • Submission of camera-ready poster: August 15, 2019
  • Poster session: August 20, 2019

Submission Rules

  • First author must be student, i.e., authors may not have received their PhD degree before August 2019.
  • Submissions may be singly or jointly authored.
  • Submissions should not be any longer than 4 pages for a poster presentation (including examples and references).
  • Submissions must be in the format provided at the SSDA website.
  • Submissions must be received by August 9, 2019 (by 23:59 GMT time)
  • Please send your proposals and posters to info@pprs.org with subject: SSDA 2019 Poster - <AuthorName>

Click here to download poster template

Programming Competition

Sketches have been the oldest mode of human communication which led to the development of symbols and alphabets and eventually formal handwriting. Sketch recognition refers to automated recognition of hand-drawn diagrams using computer programs. The key challenge in computerized recognition of sketches is the unique artistic style of an individual leading to high intra class variations. The competition involves developing a sketch recognition system using any of the deep learning based techniques.

Dataset

To train and evaluate the system, TU-Berlin sketch dataset presented in [1] will be employed. Sample sketches from the database are illustrated in Figure 1.

Figure 1 Sample sketches from the database

The database comprises 250 classes of objects and each class has 80 instances. For this particular task, the database is divided into three subsets as follows.

  • Training Set: 40 images per class – 40 x250 = 10,000 images
  • Validation Set: 20 images per class – 20 x 250 = 5,000 images
  • Test Set: 20 images per class – 25 x 250 = 5,000 images

Competition Protocol & Evaluation Metric

The training and validation datasets (with ground truth) will be made available on Day 1 of the competition. The test data will be made available on the final day of competition. Participants are free to use any of the exiting techniques or implementations or come up with their own innovative solutions. It is highly recommended to save the model and weights and not to re- train the model once the test data is provided. Only one system can be submitted by each participant and the systems will be ranked using classification rate as the evaluation metric.

Submission Rules

Participants need to submit predictions on test data in the form of a CSV file containing query image name and corresponding class labels (similar to ground truth of training and validation datasets). Winner will be announced in the closing session of the school.

References

[1] Eitz, Mathias, James Hays, and Marc Alexa. "How do humans sketch objects?." ACM Trans. Graph. 31.4 (2012): 44-1.

Dataset can be found here

Competition Test Data and Submission of Results

Test data for project competition is available here.

Submit your CSV files in the same format as that of training/validation data (i.e. image_name, class label) by Friday 10:00 AM.

Speakers

Cheng-Lin Liu

 CASIA, China

Seiichi Uchida

 Kyushu University, Japan

Marcus Liwicki

 Luleå, Sweden

Adrian Ulges

 RheinMain University of Applied Sciences, Germany

Ulrich Schwanecke

 RheinMain University of Applied Sciences, Germany

Sheraz Ahmed

 DFKI, Germany

Organizers

Dr. Faisal Shafait

 Professor
National University of Sciences and Technology, Pakistan

Dr. Imran Siddique

 Professor
Bahria University, Pakistan

Dr. Khurram Khurshid

 Professor
Institute of Space and Technology, Pakistan

Dr. Adnan Ul-Hasan

 Research Lead
Deep Learning Lab, National Center of Artificial Intelligence, Pakistan

Dr. Imran Malik

 Assistant Professor
National University of Sciences and Technology, Pakistan

Dr. Shiekh Faisal Rashid

 Assistant Professor
University of Engineering and Technology, Lahore Pakistan

Nosheen Abid

Research Associate
National University of Sciences and Technology, Pakistan

Jaleed Khan

Research Associate
Institute of Space and Technology, Pakistan

Association

Supported By

Organized By

Collaborators

How To Reach The Venue

By Train -- Islamabad Railway Station

You can take a cab or book Uber/Careem to reach NUST. It is 15 min (11 Km) from the Margalla Railway Station, Islamabad.

By Road -- From M1/M2 Motorways

The main entrance of NUST is on Kashmir Highway, which is connected to M1/M2 Motorway on a signal-free corridor.

By Air -- From Islamabad International Airport

Islamabad International Airport is at a drive of 1 hour from the event venue. Shuttle service from the airport to the event venue/major hotels will be provided to the participants. Please provide your arrival details to the organizing committee at least 48 hours before your arrival.

Map

School of Electrical Engineering and Computer Science
National University of Sciences and Technology
Sector: H-12, Islamabad

Accomodation

There are several hotels in Islamabad where participants can get accommodation such as Islamabad Serena Hotel (from $275), Ramada Hotel (from $150), Islamabad Club, and Best Western Hotel (from $70). Furthermore, on-campus accommodation will also be available at NUST (from $30). NUST has numerous boarding facilities for students as well as academicians visiting from abroad.

Get in touch

Phone

(92)51 90852064

Email

info@pprs.org.pk

Address

134 Street 9, I-10/3 Islamabad, Pakistan 44000

Facebook

facebook link
Registration
Early Bird
(till June 15, 2019):
Student: PKR 3,500/- (USD 25/-)
Professional: PKR 15,000/- (USD 110/-)
Regular Registration
(till July 31, 2019):
Student: PKR 5,000/- (USD 35/-)
Professional: PKR 20,000/- (USD 145/-)
Late Registration
(till August 17, 2019):
Student: PKR 7,500/- (USD 55/-)
Professional: PKR 25,000/- (USD 180/-)