{"id":"58fbf256-97a9-486a-8c23-844a8b8e64fc","arxiv_id":"2412.14631","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A structured review of 158 fruit tree segmentation papers argues that the main bottleneck is the absence of versatile datasets and models reusable across agricultural tasks.","lead":"This paper reviews 158 studies on segmenting fruit trees in front-view images, organizing them by method, image type, agricultural task, and fruit species. It proposes a 'crawling review' method and concludes that the field lacks versatile datasets and models that transfer across tasks and orchards.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"A single-seed, three-journal crawl cannot support an absence claim: if the 158-paper corpus under-covers front-view fruit-tree segmentation, the 'lack of a versatile dataset/model' finding is unsupported. Keep the verdict conditional until recall is demonstrated.","rationale":"I agree with the reader that corpus completeness is the weakest assumption, and I find it genuinely load-bearing rather than a methodological nicety: the headline is a negative claim, and negative claims inherit all recall limitations of the search that produced them. The specific aggravating factors are the domain mismatch of the single seed paper ([Chehreh 2023] is UAV/top-view/forestry-oriented) and the narrow supplementary window (three journals, 2020–2023), which leave recent front-view papers in other venues reachable only via citation chains. The manuscript itself flags its epistemic limits: Section 2.2.1 asserts exhaustiveness without evidence, and Section 5.2 supports the versatility gap only 'to the best of our knowledge.' I am not raising a consensus-based objection; the concern is internal robustness of an absence argument. I would credit the paper for enumerating the corpus in Tables 1–2 and for the dataset appendix with URLs, which make a recall audit feasible—and that feasibility is exactly why the audit should be the acceptance condition. Because my concern matches the reader's weakest assumption and does not move the verdict, CONDITIONAL stays as the appropriate assessment.","tokens_in":41977,"tokens_out":6872,"duration_ms":48065,"concrete_test":"Run a documented recall audit: query Scopus or Web of Science over 1990–2023 with fruit/tree/orchard terms, segmentation terms, and the taxonomy's task terms (harvesting, phenotyping, pruning, spraying, thinning, yield, navigation), compare the retrieved set against the 158 papers in Tables 1–2, and inspect the retrieved-but-missing papers. If any missing paper proposes a multi-species dataset (3+ fruit types) or a single segmentation model evaluated on more than one task/environment, the 'lack of a versatile dataset and model' conclusion is undermined. Secondary probe: re-run the same crawling procedure from an independent seed (e.g., [Xiao 2023] or [Farjon 2023]) and measure corpus growth; growth beyond roughly 10% would show the original crawl was not exhaustive, so the CONDITIONAL verdict should remain until the recall audit is published.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The review's central claim—that previous studies most notably lack a versatile dataset and segmentation model—is an absence claim, and its entire evidence base is the crawling corpus of 158 papers (Section 2.2.1). The crawl starts from one seed paper, [Chehreh 2023], which is a UAV/top-view, digital-forestry-oriented survey, and the supplementary phase scans only three journals (Computers and Electronics in Agriculture, Biosystems Engineering, Journal of Field Robotics) for January 2020–December 2023. Consequently, front-view agricultural-domain papers published in venues outside those three journals (e.g., Precision Agriculture, Sensors, Remote Sensing, Agronomy, IEEE Access/RAL, Frontiers in Plant Science) can enter the corpus only through citation chains. Papers from 2022–2023 that have not yet been cited, or older papers not in the seed's citation ancestry, are systematically excluded. The paper asserts (Section 2.2.1) that crawling is 'closer to an exhaustive search,' but no saturation analysis, no independent query cross-check, and no reproducible protocol is provided; inclusion decisions (front-view vs top-view, fruit vs forest) are applied manually and are not operationalized. The manuscript itself flags the fragility: Section 5.2 supports the versatility-gap finding only 'to the best of our knowledge,' citing [Siddique 2022] as the sole multi-species example. Credit is due for enumerating the corpus in Tables 1–2 and providing dataset URLs in Table A.1, but enumeration is not completeness. If even a few missed papers report cross-species datasets or single models evaluated across tasks/environments, the headline deficiency conclusion is weakened.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript surveys front-view fruit tree image segmentation research published between 1990 and 2023. The author introduces a 'crawling review' method: starting from the single seed paper [Chehreh 2023], the bibliography is expanded through citation chains and supplemented by scanning three named journals for 2020-2023, yielding 158 papers. The papers are organized through a taxonomy of method (rule-based vs. deep learning), image type (RGB, RGB-D, point cloud, other), agricultural task (phenotyping, harvesting, spraying, pruning, etc.), and fruit species, and each paper is summarized in Sections 3-4. Statistics on publication date, image type, task, and fruit are presented in Figure 5. The paper's central claim is that the most significant shortcoming in prior work is the absence of versatile datasets and segmentation models that transfer across tasks and environments, and it lists six future research directions. The appendix inventories 11 public datasets and provides background on segmentation methods, performance metrics, and agricultural tasks.","tokens_in":42271,"tokens_out":6164,"duration_ms":44896,"significance":"If the corpus were complete, the paper would provide a useful first synthesis of front-view fruit tree segmentation, an important niche between agricultural robotics and computer vision. The author deserves credit for enumerating and narratively summarizing 158 papers (Tables 1-2), making dataset URLs available in Table A.1, and giving a clear taxonomy that practitioners can follow. The discussion of task-specific sensor choices and the six future directions are sensible. However, the paper's central contribution is an absence claim about the field ('lack of a versatile dataset and segmentation model'), and that claim currently rests on a non-validated manual crawl. The value of the review therefore depends on either making the corpus demonstrably complete or explicitly reframing the conclusion as a property of the collected corpus.","major_comments":[{"comment":"The central absence claim in the Abstract and Section 5.3 ('the most noticeable deficiency ... lack of a versatile dataset and segmentation model') is not supported by the evidence presented for the corpus. The crawling phase starts from a single seed paper [Chehreh 2023], which is a UAV/top-view digital-forestry-oriented survey, and the supplementary phase scans only Computers and Electronics in Agriculture, Biosystems Engineering, and Journal of Field Robotics for 2020-2023. Front-view agricultural papers in venues such as Precision Agriculture, Sensors, Remote Sensing, Agronomy, IEEE Access/RAL, and Frontiers in Plant Science can enter the corpus only if they are cited by the seed or its citation descendants, so recent or less-cited papers are systematically excluded. The manuscript states that crawling is 'closer to an exhaustive search' but provides no saturation curve, no recall comparison against a database query, and no reproducibility package. I recommend either adding a validation study that demonstrates recall (e.g., a second crawl from multiple seeds or a WoS/Scopus query with overlap analysis) or softening the conclusion to describe what is observed in the 158-paper corpus, not in the field as a whole.","section":"§2.2.1, Abstract, §5.3"},{"comment":"The taxonomy and the numerical statistics in Figure 5 depend on manual inclusion decisions (front-view vs top-view, fruit tree vs forest tree, rule-based vs deep learning, task label, fruit label), but the manuscript does not report operational definitions for these decisions, a second annotator, or an inter-rater reliability check. Because Tables 1-2 and Figure 5 are the basis for the qualitative trends in Sections 3-5, misclassifications at this stage propagate into the conclusions. Please provide explicit inclusion/exclusion criteria and at least a small reliability study, or present the tables and statistics as illustrative rather than exhaustive.","section":"§2.2.2, §2.3, Tables 1-2"},{"comment":"The versatility-gap conclusion is not operationalized. Section 5.2 supports the claim that multi-species or multi-task models are essentially absent by citing [Siddique 2022] as the single example 'to the best of our knowledge,' and Section 5.3 supports the dataset-deficiency claim with the 11 datasets in Table A.1. The review does not define what counts as 'versatile' (e.g., number of tasks, number of species, varied illumination/season/architecture), nor does it systematically evaluate each of the 158 papers against that criterion. Without such a criterion, the 'most noticeable deficiency' finding cannot be distinguished from a general impression. I suggest adding a small systematic table or analysis that checks each corpus paper for multi-task/multi-environment evaluation, which would turn this claim into a verifiable statement.","section":"§5.2-§5.3, Table A.1"}],"minor_comments":[{"comment":"The text refers to 'Silwal et al.' for the apple-picking robot, but the reference list entry is [Siwal2017]; please unify the spelling (Abhisesh Silwal).","section":"§3.1.2 / References"},{"comment":"In the paragraph on AlexNet, 'won first place Krizhevsky 2012]' is missing the opening bracket; it should read '[Krizhevsky 2012].'","section":"Appendix A.1.3"},{"comment":"'tomato and maze' and 'tomatoes and maze' should read 'maize' in both occurrences.","section":"§4.3.1 and A.3.3"},{"comment":"In the description of [Hung 2013], 'conditional random file' should be 'conditional random field (CRF).'","section":"§4.4.2"},{"comment":"For reproducibility, specify whether the citation-crawling phase had any date restriction, how citation chains were followed (forward citations, backward citations, or both), and how duplicate or inaccessible papers were resolved.","section":"§2.2.1"}],"recommendation":"major_revision","confidential_remarks":"The main issue is that the paper's headline conclusion is an absence claim; the author should be pushed to either add a recall-validation appendix or to rephrase the conclusion as corpus-relative. I do not see a mathematical/technical circularity problem, and the individual paper summaries appear faithful to the cited abstracts, so the path to acceptance is through methodological transparency rather than a change of topic."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, if you work on agri-vision, this is worth a skim: it is the first review I know of that specifically covers front-view fruit tree segmentation for agricultural tasks, with a structured taxonomy and 158 summarized papers. The tables by image type/task/fruit are practical, and the appendix listing 11 public datasets with URLs is genuinely useful. It also documents the RB-to-DL shift nicely. So the paper earns its place as a field map.\n\nThe soft spot is the crawling review method. It starts from one seed (Chehreh 2023, which is UAV/top-view forestry) and supplements with only three journals from 2020–2023. That means papers in other venues—Precision Agriculture, Sensors, RAL, etc.—enter only via citation chains. The paper says the crawl is 'closer to an exhaustive search,' but no saturation analysis or reproducible protocol is given, and inclusion decisions are manual with no inter-rater check. The consequence is that the central claim—that previous studies most notably lack a versatile dataset and segmentation model—is an absence claim resting on a possibly incomplete corpus. The paper itself hedges: in Section 5.2 it says 'to the best of our knowledge' the only multi-species example is [Siddique 2022]. That is honest but thin. If a few missed papers report cross-species models or multi-task datasets, the headline finding weakens.\n\nI don't think this sinks the paper. The review's descriptive content—the summaries, taxonomy, and dataset list—stands on its own. The deficiency claim is plausible and the author flags it as a knowledge claim rather than a proven fact. But for the paper to be accepted as the definitive map, the author should either release the full corpus list with inclusion criteria, or at least do a cross-check against a database query to show recall. That would turn the conditional into something stronger.\n\nWho is this for? Masters students and researchers entering orchard robotics who need a quick orientation. It also serves as a case study in why snowball-based reviews need a recall check before making absence claims. I'd accept it for peer review, with a request to address corpus completeness. I would cite it with a caveat.","headline":"Useful field map of front-view fruit-tree segmentation, but the headline deficiency claim leans on a corpus that may not be complete; keep the verdict conditional.","tokens_in":42811,"tokens_out":2651,"would_cite":true,"duration_ms":21717,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This review of 158 papers finds that fruit tree image segmentation is dominated by task- and environment-specific solutions, and that the field's most noticeable deficiency is the lack of a versatile dataset and segmentation model.","keywords":["fruit tree segmentation","front-view images","crawling review","rule-based segmentation","deep learning segmentation","precision agriculture","public datasets","versatile model"],"falsifier":"A comprehensive keyword search of the published literature on front-view fruit tree segmentation, followed by checking whether every result falls inside the 158-paper corpus or is legitimately excluded as fruit-only, forest, or top-view work, would test the review's completeness. Finding even one pre-2023 study that supplies a public dataset or a model used across several agricultural tasks and environments would directly weaken the claim that no versatile dataset or segmentation model exists.","tokens_in":41756,"feed_emoji":"🍎","tokens_out":7030,"duration_ms":47140,"temperature":0.7,"pith_summary":"This review brings together 158 papers on segmenting front-view images of fruit trees, the perspective that harvesting, spraying, pruning, and monitoring machines actually see. The papers are organized by a taxonomy that goes from method (rule-based versus deep learning) to image type, agricultural task, and fruit species. The central claim is that the most noticeable deficiency of previous studies is the lack of a versatile dataset and segmentation model that can be applied across a variety of tasks and environments. A sympathetic reading of the review is that the field has accumulated many task-specific solutions, each with its own private data, and that this fragmentation blocks objective comparison and slows the deployment of agricultural robots. The paper closes by proposing six future research tasks aimed at a single versatile tree segmentation module.","feed_headline":"No versatile model exists for fruit tree segmentation","feed_subtitle":"A front-view review of 158 papers maps task-specific solutions and six roads to a general module.","key_machinery":"The paper's organizing device is a four-level taxonomy that classifies every collected paper in the fixed order method (rule-based versus deep learning), image type (RGB, RGB-D, point cloud, others), agricultural task (phenotyping, harvesting, spraying, pruning, yield estimation, navigation, thinning, training), and fruit species. This taxonomy is what turns the 158-paper corpus into a map of the field, exposing the dominance of deep learning, the concentration on apples and grapes, and the task-specific fragmentation that underlies the deficiency claim. The companion mechanism is the crawling review itself, a citation-following search procedure analogous to web crawling: it starts with a seed paper, pushes cited papers into a queue, processes them until the queue is empty, and optionally adds recent papers from selected journals. Together these two mechanisms define both the evidence base and the viewpoint from which the review reads the literature.","core_discovery":"The paper presents itself as the first review of front-view fruit tree segmentation in the agricultural domain, departing from earlier tree-segmentation surveys oriented to top-view UAV images and digital forestry. Using a crawling review that begins with a seed paper, follows citations until the queue is empty, and supplements with recent issues of three journals from 2020 to 2023, it collected 76 rule-based and 82 deep-learning papers from 1990 to 2023. Classified by method, image type, agricultural task, and fruit, the corpus shows a paradigm shift: rule-based papers peaked around 2015-2018 and then declined, while deep-learning papers appeared around 2018 and kept increasing, with RGB images becoming dominant and harvesting overtaking phenotyping as the most frequent task. The review's central conclusion is that no versatile dataset and no versatile segmentation model exist: the 11 public datasets it lists are each highly specific to one task or environment, so performance results from different papers cannot be objectively compared. It therefore recommends building versatile datasets and models, using few-shot and self-supervised learning, fusing CNNs with transformers, monocular depth estimation, and DL-based 3D reconstruction as routes to a general tree segmentation module.","pith_inferences":["If the fragmentation thesis is correct, then building a multi-task, multi-environment benchmark should be the field's first priority; such a benchmark would likely be harder than any existing dataset and would expose current model limitations.","The review's public-dataset inventory suggests that private task-specific datasets are the norm; a coordinated effort to publish and standardize them could be as valuable as any single algorithmic advance.","The crawling review's completeness could be tested by repeating it from several different seed papers; if the corpus grows substantially, the statistics and the deficiency claim might need revision.","The proposed CNN-transformer fusion and monocular depth estimation are concrete, testable next steps: one could directly compare fused versus pure-CNN models on thin branches and occluded fruit, which the review identifies as hard cases."],"forward_implications":["If the field lacks a versatile dataset and model, then each new agricultural task or environment requires designing, training, and testing a new method, which is a major barrier to applying computer vision broadly in orchards.","Because there is no standard dataset, objective performance comparison between segmentation methods is currently of little value; a shared benchmark would change that.","The deep-learning era favors cheap RGB and RGB-D sensors; point clouds and multi-spectral images are increasingly rare, so future systems will likely build on smartphone cameras and low-cost depth sensors.","Few-shot and self-supervised learning are identified as ways to overcome the scarcity of labeled agricultural data and to move toward a versatile model.","A versatile dataset, built with horticultural expertise, could act as a de facto standard and motivate challenges that drive the field forward."],"supporting_citations":[{"why":"serves as the seed paper for the crawling review and as the prior tree-segmentation survey oriented to UAV top-view images that this review defines itself against.","marker":"[Chehreh 2023]"},{"why":"supplies the systematic-review methodology that the crawling review contrasts with and extends.","marker":"[Snyder 2019]"},{"why":"surveys public datasets for computer vision in precision agriculture and is cited as evidence that existing agricultural datasets are deficient in quantity and quality.","marker":"[Lu 2020]"},{"why":"surveys computer vision in urban and controlled-environment agriculture and supports the paper's assessment of dataset and application gaps.","marker":"[Luo 2023]"},{"why":"provides the Uni-perceiver example of a unified versatile model in AI that motivates the call for a versatile tree segmentation module.","marker":"[Zhu 2022]"},{"why":"demonstrates a multi-species flower segmentation model and contributes the public Fruit Flower dataset, one of the few cross-species generalization attempts in the corpus.","marker":"[Dias 2018a]"},{"why":"extends flower segmentation with self-supervised contrastive learning across apple, peach, and pear, underpinning the review's future-work direction on self-supervised learning.","marker":"[Siddique 2022]"},{"why":"contributes the public NIHHS-JBNU dataset of intertwined apple trees and exemplifies both the task-specific nature of existing datasets and the hard problem of adjacent-tree boundaries.","marker":"[La 2023]"},{"why":"is the fruit-only segmentation survey used to define the review's scope boundary by excluding fruit-only work.","marker":"[Xiao 2023]"}],"fun_headline_variants":["158 papers, zero universal fruit tree model","Fruit tree segmentation: no one-size-fits-all model","Every fruit tree model is task-specific, review finds","Deep learning dominates, yet no general fruit tree model","Six future tasks to build a versatile fruit tree model"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The review's load-bearing premise is that its crawling search, starting from a single seed paper and supplementing with three journals from 2020 to 2023, captured essentially all relevant front-view fruit tree segmentation work; if a sizable body of such work lies outside that citation graph, the statistics and the deficiency conclusion could misrepresent the field.","fun_headline_variants_meta":{"raw":{"variants":["158 papers, zero universal fruit tree model","Fruit tree segmentation: no one-size-fits-all model","Every fruit tree model is task-specific, review finds","Deep learning dominates, yet no general fruit tree model","Six future tasks to build a versatile fruit tree model"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00035,"raw_usage":{"total_tokens":1900,"prompt_tokens":922,"completion_tokens":978,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":902}},"tokens_in":538,"tokens_out":978,"duration_ms":6166,"temperature":1.0,"reasoning_tokens":902,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:02:39.166564+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A comprehensive keyword search of the published literature on front-view fruit tree segmentation, followed by checking whether every result falls inside the 158-paper corpus or is legitimately excluded as fruit-only, forest, or top-view work, would test the review's completeness. Finding even one pre-2023 study that supplies a public dataset or a model used across several agricultural tasks and environments would directly weaken the claim that no versatile dataset or segmentation model exists.","supporting_citations":[],"review_version":1}