Challenge data

Description

Competitive challenge

Industrial

Classification

Images

10MB to 1GB

Intermediary level

Dates

Started on Aug. 1, 2025

Challenge context

Challenge goals

Data description

The training dataset consists of 8,278 images, divided into 6 classes:

1 class for good parts.
5 classes for defects.
The classes are highly imbalanced:

GOOD: (Label = 0) 				**1,235 images**
Flat loop: (Label = 1) 			**71 images**
White lift-off: (Label = 2) 	**270 images**
Black lift-off: (Label = 3)	**104 images**
Missing: (Label = 4) 			**6,472 images**
Short circuit MOS: (Label = 5)	**126 images**
The test dataset contains 1,055 images divided into 7 classes:

1 class for good parts.
5 defect classes (same as in the training set).
1 “drift” class (Label = 6), which includes various types of images that do not belong to any other class. This class represents anomalies in image capture (e.g., blurry images) or certain rare defects grouped together to form this anomalous class.
For both datasets, there is diversity in the captured images, indicated in the columns ‘window’ and ‘lib’ of the file output_train.csv:

The ‘lib’ indicator represents one of the 4 different components present in the parts. Their orientation relative to the camera and their size vary.
The ‘window’ indicator represents one of the 2 inspection zones of the component.

Benchmark description

Files

Files are accessible when logged in and registered to the challenge

The challenge provider

PROVIDER WEBSITE

Challenge Data

Improving Industrial Quality Control with Computer Vision
by Valeo

Description

Dates

Challenge context

Challenge goals

Data description

Benchmark description

Files

The challenge provider

Challenge Data

Improving Industrial Quality Control with Computer Vision by Valeo

Description

Dates

Challenge context

Challenge goals

Data description

Benchmark description

Files

The challenge provider

Improving Industrial Quality Control with Computer Vision
by Valeo