DAssd-Net: A Lightweight Steel Surface Defect Detection Model Based on Multi-Branch Dilated Convolution Aggregation and Multi-Domain Perception Detection Head
Bibliographic record
Abstract
During steel production, various defects often appear on the surface of the steel, such as cracks, pores, scars, and inclusions. These defects may seriously decrease steel quality or performance, so how to timely and accurately detect defects has great technical significance. This paper proposes a lightweight model based on multi-branch dilated convolution aggregation and multi-domain perception detection head, DAssd-Net, for steel surface defect detection. First, a multi-branch Dilated Convolution Aggregation Module (DCAM) is proposed as a feature learning structure for the feature augmentation networks. Second, to better capture spatial (location) information and to suppress channel redundancy, we propose a Dilated Convolution and Channel Attention Fusion Module (DCM) and Dilated Convolution and Spatial Attention Fusion Module (DSM) as feature enhancement modules for the regression and classification tasks in the detection head. Third, through experiments and heat map visualization analysis, we have used DAssd-Net to improve the receptive field of the model while paying attention to the target spatial location and redundant channel feature suppression. DAssd-Net is shown to achieve 81.97% mAP accuracy on the NEU-DET dataset, while the model size is only 18.7 MB. Compared with the latest YOLOv8 model, the mAP increased by 4.69%, and the model size was reduced by 23.9 MB, which has the advantage of being lightweight.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".