Bottle-SKU benchmark and ProDMM target shampoo-bottle defects
Original title: Prototype-Guided Dynamic Domain Modulation and a Large-Scale Multi-SKU Benchmark for Shampoo Bottle Surface Defect Detection
The dataset has 13,119 images, 20,419 boxes, seven defect types, and 48 bottle SKUs.
The authors release Bottle-SKU, a multi-SKU benchmark for shampoo-bottle surface defect detection that pairs instance boxes with product-variant identities. Prototype-Guided Dynamic Domain Modulation builds an SKU visual prototype, predicts a soft SKU distribution, and applies low-rank Query/Value modulation plus residual conditioning while keeping the pretrained Grounding DINO pathway. Training combines detection, SKU-domain classification, and modulation regularization. Experiments on Bottle-SKU and public GC10-DET-SKU are reported as evidence that prototype-guided conditioning helps under SKU appearance shifts.
Unconfirmed: The abstract says experiments demonstrate effectiveness but does not report accuracy numbers.
- Bottle-SKU covers geometry, material, label, reflectance, and imaging differences.
- ProDMM preserves the pretrained Grounding DINO detection pathway.
- Evaluation also uses the GC10-DET-SKU metallic surface dataset.
- No numeric accuracy figures are stated in the abstract.
- AI models
- ProDMMGrounding DINO
From the original article
1 min read at arxiv.orgIndustrial shampoo-bottle surface defect detection must operate across many stock keeping units (SKUs), where differences in bottle geometry, material, label layout, reflectance, and imaging conditions can substantially alter the appearance of the same defect.
By Yuehang Li · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.