TILDA-OBB
Description
The annotated dataset, TILDA-OBB, contains 1,562 images across four fabric groups: Plain (C1), Texture (C2), Printed (C3), and Patterned (C4). It is derived from the TILDA Textile Texture Database (Schulz-Mirbach, 1996), developed at Technische Universität Hamburg-Harburg under the DFG research programme "Automatic Visual Inspection of Technical Objects." TILDA contains 3,200 grayscale images (768×512) across eight fabric types and seven defect classes, plus one defect-free class (Albert-Ludwigs-Universität Freiburg, 1995). The raw images carry only short German text labels with no bounding-box annotations; these were translated into English and annotated in Roboflow (Dwyer and Nelson, 2024) using oriented bounding boxes. Five defect classes were retained: Cut, Foreign Object, Hole, Stain, and Thread Defect. The No Defect and Wrinkle classes were excluded, as neither contains a spatially distinguishable defect region.
Files
Institutions
- University of Global Village (UGV)Barisal Division, Barishal