Industrial Visual Quality Inspection Challenge
Join the challenge for detecting defects in real-world manufacturing conditions
JOIN THE IVQI CHALLENGE
In many real-word scenarios, it is very difficult and time-consuming to collect training data with all defect types as new defects may occur during production. Due to the lack of all possible defects in the training dataset, anomaly detection is often performed by models trained via unsupervised learning or supervised learning using data with synthetic defects or anomalies.
Part of the workshop will be a challenge with a newly introduced dataset (the dataset will consist of approximately 50 product types from 10 manufacturing domains consisting roughly of 10 000 fully annotated pictures).
JOIN THE IVQI CHALLENGE
In many real-word scenarios, it is very difficult and time-consuming to collect training data with all defect types as new defects may occur during production. Due to the lack of all possible defects in the training dataset, anomaly detection is often performed by models trained via unsupervised learning or supervised learning using data with synthetic defects or anomalies.
Part of the workshop will be a challenge with a newly introduced dataset (the dataset will consist of approximately 50 product types from 10 manufacturing domains consisting roughly of 10 000 fully annotated pictures).
NEW METHODS FOR AUTOMATIC QUALITY CONTROL
The dataset will simulate complex conditions that can be encountered in real-world visual inspection applications, e.g. metal products, painted components, fabric structures, food packaging, etc. The main purpose of the workshop will be to encourage the development of new methods that can be applied in the field of automatic quality control. More specifically, in real manufacturing conditions where current methods do not provide sufficient precision. The workshop will be mainly aimed (but not limited) to cover the following topics:
- Designing new Data-centric approaches to prepare smart data rather than big data.
- Development of new methods to detect image-level and pixel-level anomalies.
- Proposing new approaches targeting small training data challenges.
- Introduction of novel approaches for generating synthetic defects and data augmentation.
- Transferring knowledge among various use cases or camera and light setups.
NEW METHODS FOR AUTOMATIC QUALITY CONTROL
The dataset will simulate complex conditions that can be encountered in real-world visual inspection applications, e.g. metal products, painted components, fabric structures, food packaging, etc. The main purpose of the workshop will be to encourage the development of new methods that can be applied in the field of automatic quality control. More specifically, in real manufacturing conditions where current methods do not provide sufficient precision. The workshop will be mainly aimed (but not limited) to cover the following topics:
- Designing new Data-centric approaches to prepare smart data rather than big data.
- Development of new methods to detect image-level and pixel-level anomalies.
- Proposing new approaches targeting small training data challenges.
- Introduction of novel approaches for generating synthetic defects and data augmentation.
- Transferring knowledge among various use cases or camera and light setups.
THE ORGANIZERS
The organizers will be namely: Pavel Dvorak, PhD (KM), Radim Burget, PhD, Martin Jonak, PhD, Stepan Jezek (all BUT), Francesca Odone, PhD, Prof. Vittorio Murino, PhD (both UNIGE).
THE ORGANIZERS
The organizers will be namely: Pavel Dvorak, PhD (KM), Radim Burget, PhD, Martin Jonak, PhD, Stepan Jezek (all BUT), Francesca Odone, PhD, Prof. Vittorio Murino, PhD (both UNIGE).
GET IN CONTACT
Please leave your message:
VISIT THE VIRTUAL TOUR OF OUR LABORATORY
Our exclusive virtual tour devoted to interactive online demos.
