Pith. sign in

REVIEW 4 major objections 5 minor 26 references

Detailed Evaluation of Modern Machine Learning Approaches for Optic Plastics Sorting

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper finds that RGB-only computer vision sorts plastics by appearance, not material identity, so real-world sorting needs extra modalities.

desk verdict Useful dataset work wrapped in an unsupported conclusion: the broad claim that RGB optic sorting fails at MRFs doesn't follow from 50 training images and untuned ImageNet models. read the letter →

arxiv 2505.16513 v1 pith:5LED75TW submitted 2025-05-22 cs.CV

Vaishali Maheshkar , Aadarsh Anantha Ramakrishnan , Charuvahan Adhivarahan , Karthik Dantu This is my paper · ORCID
classification cs.CV
keywords plasticsortingopticalrecognitiondeeplearningGrad-CAMinstancesegmentationmaterialsrecoveryfacilityclassificationmultimodalsensing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Using 20,000+ images from clean catalogs, web sources, consumer-packaging photos, and a working materials recovery facility, the paper trains standard deep-learning detectors and classifiers and inspects where their attention falls. It finds that models key on physical appearance—object shape, color, packaging labels, and even conveyor-belt background—rather than on the material itself, so accuracy drops sharply from clean images to real-world streams. The paper's central claim is that optic (RGB-image) recognition methods have limited success in accurate sorting of real-world plastics at MRFs precisely because they rely on such physical properties. If true, this reframes the sorting problem: reliable resin identification needs additional modalities, such as near-infrared spectroscopy, fused with vision.

What carries the argument

The argument is carried by pairing object-detection and classification models with attribution visualizations: Mask R-CNN for instance segmentation, ResNet-34, ResNet-50, and Xception for classification, and Grad-CAM and intermediate feature maps to localize the image regions that drive each prediction. Grad-CAM (gradient-weighted class activation mapping) produces heatmaps of the pixels most responsible for a classification, and those heatmaps are the evidence that the model is keying on conveyor belts, labels, or shape rather than polymer identity.

What would settle it

Take a single mixed-plastics conveyor, sort with an RGB-only detector, a near-infrared sensor, and a fused system, and audit ground-truth resin codes by hand: if the RGB-only detector matches the NIR sensor on deformed and contaminated items, the paper's central claim fails. Alternatively, retrain on images with labels and colors removed; high accuracy with attention on the plastic body would contradict the claim that models rely on those cues.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is an empirical diagnosis of why modern optical sorting models underperform on real materials-recovery streams. Trained models reach roughly 80% mean average precision on clean factory-catalog images and about 96% classification accuracy on clean web-scraped plastics, but accuracy falls to 71.8% on consumer-packaging photos, and the MRF data expose the mechanism: Grad-CAM and feature maps show the network attending to conveyor belts, beverage labels, metal cans nearby, and shape outlines instead of the plastic polymer surface. From this the authors conclude that optic recognition, as embodied by these RGB-based CNN models, does not reliably identify plastic type in contaminated, deformed, mixed streams, and they recommend combining vision with spectroscopic modalities.

Load-bearing premise

The load-bearing premise is that the RGB-image CNNs trained here represent 'optic recognition' generally, so that attention maps focusing on conveyor belts, labels, and shape are enough to show optical sorting is unreliable; the paper does not test NIR sensors or measure end-to-end sorting accuracy.

Editorial extensions

If this is right

  • Clean-dataset accuracy claims for plastic sorters should not be read as real-world performance; the same models lose more than 20 percentage points on consumer-packaging images and focus on background cues on MRF footage.
  • Any RGB-only sorting deployment at a materials recovery facility should be validated with attribution analysis before being trusted for resin-type decisions.
  • Spectroscopic sensors (near-infrared, Raman, FTIR) become the natural complement; vision can localize objects while the spectral channel identifies material.
  • The compiled multi-source image collection gives later work a concrete benchmark for measuring whether a model classifies by material identity rather than by appearance.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial — the same attention-shortcut diagnosis likely applies to other recycling domains (glass color sorting, e-waste) where visual appearance only weakly correlates with material value.
  • Editorial — a quantitative test the authors did not run is to compare end-to-end sorting accuracy of an RGB-only CNN against a co-located NIR sensor on the same conveyor; the attention maps predict RGB will lose.
  • Editorial — the paper's evidence is consistent with, but does not prove, the stronger claim that no amount of RGB training data can close the gap, because resin chemistry is not visually encoded; a training-data scaling study would be needed to test that.
  • Editorial — if recycling facilities already use NIR for resin typing, the practical implication is not to replace optics but to use vision for object localization and NIR for material classification.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper curates a collection of more than 20,000 images of plastics from web sources, a plastics manufacturer's catalog, the Open Food Facts database, and a local Materials Recovery Facility (MRF). It applies several modern computer-vision models—Mask R-CNN, ResNet-34, Xception, and ResNet-50—to classify or segment plastics, and interprets their behavior using Grad-CAM, feature maps, and confusion matrices. The central claim, stated in the abstract and Section 6, is that optic recognition methods have limited success in accurate sorting of real-world plastics at MRFs because they rely on physical properties such as color and shape, and the paper recommends multi-modal sensing as a remedy.

Significance. If established, the claim that RGB-based optical sorting is fundamentally limited at MRFs would be of considerable practical importance, motivating investment in multi-modal or spectral approaches. The compiled datasets could also serve as a useful community resource. However, the experiments as presented do not support the broad conclusion: the MRF evaluation relies on an extremely small training set and on off-the-shelf pretrained models that were never fine-tuned, no end-to-end sorting accuracy is measured, and several results (80% mAP on the US Plastics dataset, 71.8% accuracy on Open Food Facts) actually indicate that optical models can perform reasonably well when training data are adequate. The paper's own acknowledged limitations—clean catalog images, a small MRF set, and manual cleaning—undercut the generalization from qualitative heatmaps to a verdict on optical sorting technology.

major comments (4)
  1. [Section 4.1.1, item 1] The Mask R-CNN experiment on the MRF dataset is trained on only 50 images and validated on 20, and the paper reports no detection or segmentation accuracy (e.g., mAP, precision/recall) for this setup. Without any quantitative performance measure, the reader cannot assess whether the model 'successfully identifies' the plastic objects, and the claim that optical recognition has limited success at MRFs rests entirely on a statistically insignificant sample with no error bars.
  2. [Section 4.1.1, items 4 and 5] The Xception and ResNet-50 Grad-CAM experiments use ImageNet-pretrained checkpoints without any fine-tuning on MRF images. The observation that these models attend to conveyor belts and metal cans is an expected symptom of domain shift, not a property of a trained optical sorter. Interpreting these heatmaps as evidence that 'optic recognition methods rely on physical properties such as color and shape' is therefore not justified; a valid test would require models trained or fine-tuned on the target task.
  3. [Section 4.1.1, item 3 and Section 4, item 3] The Open Food Facts result (71.8% accuracy with ResNet-34) is obtained on manually cleaned images of food packaging, and the paper itself acknowledges that this dataset is limited to food products and contains no deformed plastics. This is not a proxy for MRF conditions, and the 71.8% accuracy is a moderate result that does not support the broad conclusion about limited success at MRFs.
  4. [Section 5 and Abstract] The central conclusion—that optic recognition methods have limited success in accurate sorting of real-world plastics at MRFs—is not supported by the presented evidence because the paper never measures end-to-end sorting accuracy, never compares with deployed NIR/optical sorting systems, and bases its qualitative interpretation on a handful of selected Grad-CAM images. In contrast, the Mask R-CNN trained on the US Plastics dataset reaches approximately 80% mAP (Section 4.1.1, item 2), which demonstrates that an optical system can perform well when sufficient training data are available; this weakens the paper's sweeping conclusion and suggests the authors should restrict their claims to the specific under-trained models they tested.
minor comments (5)
  1. [Author affiliations] The affiliation 'NIT T richy' contains a typo; it should be 'NIT Tiruchirappalli' or similar.
  2. [References] Reference [8] is mis-cited: the entry 'W. Yoshida, R. J. Dolan, and K. J. Friston, Game theory of mind' does not correspond to SPP-net; the proper SPP-net citation (He et al., 2015) appears to be missing.
  3. [Section 4.1.1, item 3] The description of the manual cleaning of Open Food Facts images is vague; please specify how images were selected and how many were discarded.
  4. [Section 4.1.1, item 1] The sentence 'We trained 70 (50 train and 20 validation) images' is ambiguous: it should state whether this is the total number of images or per class, and it should list the number of plastic categories.
  5. [Throughout] The paper uses inconsistent spellings such as 'Grad-CAM' and 'Grad CAM'; please standardize the terminology.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical evaluation, not a derivation whose predictions reduce to fitted inputs or self-citations.

full rationale

The paper performs an empirical evaluation rather than a derivation chain. It compiles datasets, trains or applies standard models (Mask R-CNN, ResNet-34, ResNet-50, Xception), and reports accuracy, mAP, confusion matrices, and Grad-CAM/feature-map inspections. No quantity presented as a prediction is constructed from a fitted parameter and then re-tested on the same data; the reported numbers are direct test-set measurements. The headline conclusion that 'optic recognition methods have limited success in accurate sorting of real-world plastics at MRFs, primarily because they rely on physical properties such as color and shape' is an interpretive generalization from observed behavior rather than a mathematically forced consequence of the inputs. Whether that interpretation is well supported is a question of experimental validity and strength of evidence, not circularity. The paper contains no load-bearing self-citations: references to AMP Robotics, Greyparrot, TOMRA, and standard method papers are external, and no uniqueness theorem or prior-work ansatz is invoked to forbid alternatives. The paper itself acknowledges important limitations, including that the web-scraped images are 'clean plastic samples without any deformations, which unfortunately are not representative of the real world,' that the MRF Mask R-CNN was trained on only 50 images, and that OFF images were manually cleaned. These limitations weigh against the strength of the broad conclusion, but they do not make any step circular. Under the specified standard, this is a normal non-finding.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper does not introduce new entities or fitted parameters. Its conclusions rest on interpretability assumptions (Grad-CAM fidelity), the sufficiency of RGB data for polymer classification, and the representativeness of a small MRF dataset. These are all domain assumptions rather than mathematical axioms.

assumptions (3)
  • domain assumption Grad-CAM maps are faithful indicators of what the network relies on for classification.
    The paper uses Grad-CAM to conclude that models focus on conveyor belts and labels rather than plastic material, but Grad-CAM is a post-hoc approximation with known limitations and can highlight regions that are not causally necessary.
  • domain assumption RGB color images contain sufficient information to infer plastic polymer type.
    The experiments only use RGB cameras, yet the conclusions extend to all 'optic recognition methods', including NIR-based systems. This conflation is a load-bearing assumption for the central claim.
  • domain assumption The small MRF dataset (50 training images) is representative of real-world MRF conditions.
    The strong negative conclusion about optical sorting at MRFs is drawn from a tiny training set, which may not capture the full variability of waste streams.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Detailed Evaluation of Modern Machine Learning Approaches for Optic Plastics Sorting." pith.science (2026). https://pith.science/paper/5LED75TW

@misc{pith2026250516513,
  author       = {Pith},
  title        = {Pith review of: Detailed Evaluation of Modern Machine Learning Approaches for Optic Plastics Sorting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5LED75TW}},
  note         = {Machine review of arXiv:2505.16513}
}
read the original abstract

According to the EPA, only 25% of waste is recycled, and just 60% of U.S. municipalities offer curbside recycling. Plastics fare worse, with a recycling rate of only 8%; an additional 16% is incinerated, while the remaining 76% ends up in landfills. The low plastic recycling rate stems from contamination, poor economic incentives, and technical difficulties, making efficient recycling a challenge. To improve recovery, automated sorting plays a critical role. Companies like AMP Robotics and Greyparrot utilize optical systems for sorting, while Materials Recovery Facilities (MRFs) employ Near-Infrared (NIR) sensors to detect plastic types. Modern optical sorting uses advances in computer vision such as object recognition and instance segmentation, powered by machine learning. Two-stage detectors like Mask R-CNN use region proposals and classification with deep backbones like ResNet. Single-stage detectors like YOLO handle detection in one pass, trading some accuracy for speed. While such methods excel under ideal conditions with a large volume of labeled training data, challenges arise in realistic scenarios, emphasizing the need to further examine the efficacy of optic detection for automated sorting. In this study, we compiled novel datasets totaling 20,000+ images from varied sources. Using both public and custom machine learning pipelines, we assessed the capabilities and limitations of optical recognition for sorting. Grad-CAM, saliency maps, and confusion matrices were employed to interpret model behavior. We perform this analysis on our custom trained models from the compiled datasets. To conclude, our findings are that optic recognition methods have limited success in accurate sorting of real-world plastics at MRFs, primarily because they rely on physical properties such as color and shape.

Figures

Figures reproduced from arXiv: 2505.16513 by the authors.

Figure 1
Figure 1. The architecture of Mask RCNN [7] with its components, a popular Deep Learning architecture used to [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Confusion matrix for the OFF dataset. Higher numbers along the diagonal indicate proper classification. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Grad CAM activations for samples from the MRF dataset, the activations show the regions highly focused [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: A sample image collected from a regional Materials Recovery Facility (MRF). Materials are typically [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Feature maps obtained from running the image sample from Figure 4 through the Mask RCNN model. The [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Another sample image from the MRF mixed plastics conveyor belt (left). Real-world images are challenging [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: What do optical DNN models prioritize to determine the plastic class? (a) Object Detection using Mask [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Grad CAM activations for OFF dataset using Resnet-34. The red regions in the image contributed highly [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 24 canonical work pages

  1. [1]

    AMP robotics recycling reimagined

    “AMP robotics recycling reimagined.” https://www.amprobotics.com/

    work page

  2. [2]

    Grey parrot waste intelligence

    “Grey parrot waste intelligence.” https://www.greyparrot.ai/

    work page

  3. [3]

    TITECH plastic sorting technology

    “TITECH plastic sorting technology.” https://www.pcne.eu/uploads/tx_etim/ 39747_tomra.pdf

    work page

  4. [4]

    Rich feature hierarchies for accurate object detection and semantic segmentation,

    R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierarchies for accurate object detection and semantic segmentation,” in 2014 IEEE Conference on Computer Vision and Pattern Recognition , pp. 580–587, 2014

    work page 2014

  5. [5]

    Fast r-cnn,

    R. Girshick, “Fast r-cnn,” in 2015 IEEE International Conference on Computer Vision (ICCV) , pp. 1440–1448, 2015

    work page 2015

  6. [6]

    Faster r-cnn: Towards real-time object detection with region proposal net- works,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal net- works,” in Advances in Neural Information Processing Systems (C. Cortes, N. Lawrence, D. Lee, M. Sugiyama, and R. Garnett, eds.), vol. 28, Curran Associates, Inc., 2015

    work page 2015

  7. [7]

    Mask r-cnn,

    K. He, G. Gkioxari, P. Dollár, and R. Girshick, “Mask r-cnn,” in 2017 IEEE International Conference on Com- puter Vision (ICCV), pp. 2980–2988, 2017

    work page 2017

  8. [8]

    Game theory of mind,

    W. Yoshida, R. J. Dolan, and K. J. Friston, “Game theory of mind,” PLOS Computational Biology, vol. 4, pp. 1– 14, 12 2008

    work page 2008

Show all 26 references
  1. [9]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” 2015

    2015

  2. [10]

    Cspnet: A new backbone that can enhance learn- ing capability of CNN,

    C. Wang, H. M. Liao, Y . Wu, P. Chen, J. Hsieh, and I. Yeh, “Cspnet: A new backbone that can enhance learn- ing capability of CNN,” in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR Workshops 2020, Seattle, WA, USA, June 14-19, 2020 , pp. 1571–1580, ...

    2020

  3. [11]

    Efficientnet: Rethinking model scaling for convolutional neural networks,

    M. Tan and Q. V . Le, “Efficientnet: Rethinking model scaling for convolutional neural networks,” inProceedings of the 36th International Conference on Machine Learning, ICML 2019, 9-15 June 2019, Long Beach, California, USA (K. Chaudhuri and R. Salakhutdinov, eds.), vol. 97 o...

    2019

  4. [12]

    You only look once: Unified, real-time object detection,

    J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 779–788, 2016

    2016

  5. [13]

    SSD: single shot multibox detector,

    W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. E. Reed, C. Fu, and A. C. Berg, “SSD: single shot multibox detector,” in Computer Vision - ECCV 2016 - 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part I (B. Leibe, J. Matas, N. Sebe, and...

    2016

  6. [14]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” 2018

    2018

  7. [15]

    UN environment programme on plastic pollution

    “UN environment programme on plastic pollution.” https://www.unep.org/interactives/beat- plastic-pollution/

  8. [16]

    EPA plastics material specific data

    “EPA plastics material specific data.” https://www.epa.gov/facts-and-figures-about- materials-waste-and-recycling/plastics-material-specific-data

  9. [17]

    National recycling strategy: Part one of a series on building a circular economy for all,

    A. of U.S. Environmental Protection Agency, “National recycling strategy: Part one of a series on building a circular economy for all,” 2021

    2021

  10. [18]

    “TOMRA.” https://www.tomra.com/en/

  11. [19]

    Procter & Gamble - HolyGrail

    “Procter & Gamble - HolyGrail.” https://us.pg.com/blogs/HolyGrail/

  12. [20]

    BASF - News Releases

    “BASF - News Releases.” https://www.basf.com/global/en/media/news-releases/2021/ 05/p-21-205 .html. Accessed: Insert Date

    2021

  13. [21]

    Classification of plastic bottles based on visual and physical features for waste management,

    L. R. Kambam and R. Aarthi, “Classification of plastic bottles based on visual and physical features for waste management,” in 2019 IEEE International Conference on Electrical, Computer and Communication Technolo- gies (ICECCT), pp. 1–6, 2019

    2019

  14. [22]

    Advancing plastic waste classification and recycling efficiency: Integrating image sensors and deep learning algorithms,

    J. Choi, B. Lim, and Y . Yoo, “Advancing plastic waste classification and recycling efficiency: Integrating image sensors and deep learning algorithms,” Applied Sciences, vol. 13, no. 18, 2023

    2023

  15. [23]

    Zerowaste dataset: Towards deformable object segmentation in cluttered scenes,

    D. Bashkirova, M. Abdelfattah, Z. Zhu, J. Akl, F. Alladkani, P. Hu, V . Ablavsky, B. Calli, S. A. Bargal, and K. Saenko, “Zerowaste dataset: Towards deformable object segmentation in cluttered scenes,” 2022

    2022

  16. [24]

    One-shot learning based classification for segregation of plastic waste,

    S. Agarwal, R. Gudi, and P. Saxena, “One-shot learning based classification for segregation of plastic waste,” CoRR, vol. abs/2009.13953, 2020

    2009 arXiv

  17. [25]

    Cross-modal generative models for multi- modal plastic sorting,

    E. R. Neo, J. S. Low, V . Goodship, S. R. Coles, and K. Debattista, “Cross-modal generative models for multi- modal plastic sorting,” Journal of Cleaner Production, vol. 415, p. 137919, 2023

    2023

  18. [26]

    Grad-cam: Visual explana- tions from deep networks via gradient-based localization,

    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, “Grad-cam: Visual explana- tions from deep networks via gradient-based localization,” International Journal of Computer Vision , vol. 128, p. 336–359, Oct. 2019. 8 About the Authors Vaishali Maheshkar ...

    2019

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.