SAMSort: Vision foundation model for sorting construction and demolition waste
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The rapid growth of construction and demolition waste (CDW) poses significant challenges to environmental sustainability and the availability of land resources, particularly in densely populated cities like Hong Kong, highlighting the urgent need for efficient waste sorting and recycling techniques. Traditional CDW sorting facilities (CWSFs) rely heavily on manual and mechanical operations, which are labour-intensive, costly, and often inefficient. Recent advances in computer vision offers new opportunities for automating CDW sorting, yet most existing models require large training datasets, substantial computational resources, and generally cannot estimate the precise area proportion of each waste category, limiting their practical deployment. To address these challenges, this paper introduces SAMSort, a novel framework that adapts the Segment Anything Model (SAM) for CDW sorting through parameter-efficient fine-tuning (PEFT). Six types of PEFT layers, including transformer adapter, attention adapter, multilayer perceptron adapter, and three variants of low-rank adaptation (LoRA Types I–III), are adopted to reduce the number of trainable parameters. In addition, the performance of all 24 possible combinations is evaluated. A new dataset collected from Tseung Kwan O (TKO) CWSF is constructed for model training and evaluation. The results show that SAMSort achieves competitive waste sorting performance, with an F1-score of 0.764 and an Intersection of Union (IoU) of 0.670 using only 1% of the parameters required for full fine-tuning. ➢ A construction waste sorting method based on the vision foundation model is developed for construction and demolition waste ➢ Six types of parameter-efficient fine-tuning layers are built for model adaptation ➢ A dataset is built based on the sorting data acquired from a sorting site at Tseung Kwan O for model training and evaluation ➢ The model achieves competitive waste sorting performance with 1% of the parameters
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it