Pith. sign in

REVIEW 4 major objections 6 minor 29 references

CLIPTime: Time-Aware Multimodal Representation Learning from Images and Text

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read CLIPTime extends CLIP so that a fused image-text embedding predicts both the fungal growth stage and a continuous timestamp, trained on synthetic time-aligned data.

desk verdict The Time-Transformer is a single-token MLP, so the paper's central claim of temporal modeling is unsupported; the synthetic-only evaluation doesn't rescue it. read the letter →

arxiv 2508.00447 v1 pith:XIZWKWMX submitted 2025-08-01 cs.CV cs.LG

Anju Rani , Daniel Ortiz-Arroyo , Petar Durdevic This is my paper
classification cs.CVcs.LG
keywords CLIPvision-languagemodelsfungalgrowthstagepredictiontimestampregressionmulti-tasklearningsynthetictime-aligneddatasetbiologicalprogressiontime-awarerepresentation
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

The paper sets out to show that a vision-language model built on CLIP can infer the temporal progression of fungal growth—predicting both the developmental stage (spore, hyphae, or mycelium) and a continuous age in hours—from an image and a short text description alone, without receiving a timestamp as input at test time. To this end, the authors introduce CLIPTime, which attaches a classification head and a transformer-based regression head to the element-wise sum of CLIP image and text embeddings, and train it on a synthetic time-aligned fungal growth dataset. They report 98.7% stage classification accuracy and mean absolute regression errors around 250–300 hours across stages. The central claim is that the semantic progression encoded in CLIP's joint embedding space is enough for the model to place a sample along the growth timeline. If correct, this would make pretrained vision-language models usable for time-aware biological monitoring without explicit temporal sensors or annotation at inference.

What carries the argument

The central object is the Time-Transformer: a lightweight transformer regression head that receives a fused embedding formed by element-wise summing CLIP's image and text embeddings, treats it as a sequence of length one, and passes it through two transformer encoder layers comprising layer normalization, multi-head self-attention, and a ReLU feed-forward network of hidden size 2048. Adaptive average pooling over the length dimension is followed by a linear layer and sigmoid activation that outputs a timestamp normalized to [0,1] via min-max scaling, with the stored min and max used to project back to real hours. This module carries the argument because it is what converts static semantic similarity into a continuous temporal estimate, while the parallel classification head and the combined loss $L_{\mathrm{total}} = L_{\mathrm{cls}} + L_{\mathrm{time}}$ force the shared embedding to be useful for both discrete stages and continuous time.

What would settle it

Collect real time-stamped images of fungi spanning spore to mycelium, feed them to CLIPTime with stage-neutral text prompts, and compare predicted hour values to the true elapsed times; if classification accuracy drops well below 98.7% or the mean absolute error rises far above the 250–300 hour range, the paper's claim that the model models biological progression would be falsified for real-world use.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that temporal dynamics can be recovered from the fused visual-textual semantics of a pretrained CLIP model, without any temporal token or explicit time input during inference. CLIPTime fuses image and text embeddings via element-wise summation, passes the fused vector through two transformer encoder layers with self-attention and a feed-forward network, and then predicts a normalized timestamp through a sigmoid head while a parallel linear head predicts the growth class. Training on the synthetic time-aligned fungi dataset with equal-weight classification and regression losses yields 98.7% test accuracy and temporally coherent predictions whose mean absolute error stays near 250–300 hours per stage. The authors interpret this as evidence that the Time-Transformer captures implicit temporal progression from the morphological and semantic patterns in the embeddings, while acknowledging that the early spore stage contains little learnable temporal signal.

Load-bearing premise

The load-bearing assumption is that the analytically constructed synthetic fungal dataset faithfully represents real fungal development, since every reported accuracy and error number comes from this dataset and no real fungal images are tested.

Editorial extensions

If this is right

  • CLIPTime simultaneously classifies the stage and regresses the timestamp from fused image-text embeddings, giving a single model that performs both categorical and continuous temporal prediction at inference without explicit time input.
  • Classification accuracy on the synthetic test set reaches 98.7%, with mycelium classified perfectly and most confusion occurring between spore and hyphae, which the authors attribute to transitional visual overlap.
  • The regression results show hyphae and mycelium predictions follow the true time trends, while spore predictions do not, indicating that later developmental stages carry more learnable temporal signal in this dataset.
  • The model's normalized timestamps can be re-projected to the original hour scale using stored min-max values, so the time output is directly interpretable in real units.

Reading between the lines

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

  • The strongest unstated next test is to run the same pipeline on real time-lapse microscopy images of fungal growth; the paper's claim would be corroborated only if the 98.7% accuracy and hour-scale errors survive that transfer.
  • Because the text prompt names the stage, the model could be partly reading the stage label from text rather than inferring time from image morphology; ablating the text branch or using stage-neutral captions would reveal how much of the temporal signal comes from the visual stream.
  • With a sequence length of one inside the Time-Transformer, self-attention has no cross-token interactions; 'temporal dependencies' here amount to learned nonlinear transformation of a single embedding, so the mechanism is better described as semantic regression than temporal sequence modeling.
  • The synthetic dataset's fixed stage boundaries and clean morphologies likely make the task easier than natural variation; a test with synthetic images augmented by noise, blur, or species diversity would give a more conservative performance estimate.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes CLIPTime, a multimodal framework that extends CLIP by fusing image and text embeddings and attaching two heads: a linear classifier for fungal growth stage (spore, hyphae, mycelium) and a transformer-based regression head that predicts a continuous timestamp. The model is trained and evaluated on a synthetic, time-aligned fungal growth dataset generated by the same authors. The paper reports 98.7% classification accuracy and claims that the Time-Transformer captures implicit temporal progression from static visual-textual semantics. The central claim is that CLIPTime effectively models biological progression and produces interpretable, temporally grounded outputs.

Significance. If the central claim were established, CLIPTime would be a useful demonstration of adapting vision-language models to temporally structured biological monitoring tasks. The paper provides an open-source implementation and dataset, which supports reproducibility. The multi-task formulation combining classification and regression is a reasonable design goal. However, the significance is currently limited by two load-bearing problems: the proposed temporal head cannot perform temporal reasoning over more than one time step, and the only evaluation is on a self-authored synthetic dataset without external validation. These issues undermine the claimed contribution to time-aware multimodal representation learning.

major comments (4)
  1. [§2.2, Eqs. (2)–(5)] The Time-Transformer reshapes the fused embedding to a sequence of length 1 and applies self-attention to this single token, with no positional encoding. Self-attention over a single token computes a softmax over one element, which equals 1, so multi-head self-attention reduces to a learned per-token affine transformation. The residual connection, layer norm, and feed-forward network then form an MLP-like pointwise map. Consequently, this module cannot capture temporal dependencies across time steps, and the claim in Section 2.2 that it 'captures implicit temporal progression' is unsupported. The architecture can at most learn a static nonlinear regression from the fused embedding to a timestamp.
  2. [§2.4 and §3] The only evaluation is on a synthetic dataset that the authors describe as 'analytically constructed to simulate clear distinctions between the developmental stages' (Section 2.4), and the dataset is introduced in the authors' own references [28] and [29]. No evaluation on real fungal images is reported. Because the generator explicitly couples stage labels with timestamps, the reported 98.7% classification accuracy and class-correlated time predictions may reflect this artificial coupling rather than biologically meaningful visual progression. The claim that CLIPTime 'effectively models biological progression' is therefore not established by the present evidence.
  3. [§3, Figures 6 and 7] The regression results are not uniformly strong: the spore class shows 'widely dispersed' predictions with 'no clear correlation with ground truth values,' as the paper acknowledges. The reported MAE values of approximately 260–300 hours are difficult to interpret because the timestamp range is not reported. Despite this, the conclusion states that the model demonstrates 'strong dual-task performance' and 'effective modeling of biological progression.' The evidence supports at most partial temporal modeling for hyphae and mycelium, and the spore results actively weaken the central claim.
  4. [§3 (entire experimental section)] No baselines or ablations are provided. There is no comparison with a simple MLP regression head operating on the same fused embedding, with CLIP zero-shot classification, with image-only training, or with text-only training. The loss weights alpha and beta are fixed to 1 (Section 2.3) without any ablation. Without these comparisons, the contribution of the multimodal fusion and the Time-Transformer cannot be assessed, and the reported accuracy may largely reflect the CLIP backbone or the synthetic dataset's simplicity.
minor comments (6)
  1. [§2.2, Eq. (6)] Equation (6) appears garbled in the manuscript; the denormalization formula should be written explicitly and consistently with the normalization in Eq. (1).
  2. [§3, Figure 4] Figure 4 lacks axis labels and a legend, making the training and validation loss curves difficult to interpret.
  3. [§2.4 and §3] The dataset statistics are not reported: the number of training, validation, and test samples, the per-class timestamp ranges, and the generation parameters. These details are needed to gauge the difficulty of the regression task.
  4. [§2.1] The text prompts used with the CLIP text encoder are not described; specifying them is necessary for reproducibility.
  5. [References [28] and [29]] The relationship between the 'Synthetic Time-Aligned Fungi Generation' dataset and the 'FungalZSL' dataset should be clarified, since both are cited as the data source.
  6. [§3, Figure 8] The qualitative analysis reports predicted timestamps for random test samples, but the corresponding ground-truth timestamps are not shown, so the reader cannot verify the claimed alignment.

Circularity Check

3 steps flagged · score 6.0 of 10

Central biological-progression claim reduces to fitting the authors' synthetic generator; the Time-Transformer is a one-token pointwise map renamed as temporal reasoning.

  1. fitted input called prediction [Section 2.4 and Section 3 (Eq. 1, Eq. 8; refs [28], [29])]
    "The dataset is analytically constructed to simulate clear distinctions between the developmental stages, aligning visual, textual, and temporal modalities. A detailed description of the dataset generation methodology is provided in [29]. ... The proposed model was trained and fine-tuned to jointly perform fungal growth stage classification and temporal age regression on a synthetically generated, time-aligned fungal dataset [28]."

    The regression target (timestamp) and the entire test distribution come from a synthetic generator created by the same authors (refs [28], [29]). The min-max bounds in Eq. 1 are also fit to this same dataset, and the MSE loss in Eq. 8 is minimized against those manufactured timestamps. The reported 98.7% accuracy and class-correlated time predictions therefore measure fit to the generator's own analytic stage/timestamp coupling, not to independently observed biological progression. Calling this 'effective modeling of biological progression' presents a fit to self-generated labels as an external prediction.

  2. renaming known result [Section 2.2, Eqs. (2)-(5), and Conclusion]
    "this fused vector is reshaped to be treated as a sequence of length 1: ... Due to this fixed-length sequence, no positional encoding is applied. ... By leveraging self-attention and feed-forward layers applied to this single embedding token, the module captures complex, non-linear relationships that map semantic evolution directly to real-valued timestamps."

    With sequence length fixed at 1, the softmax in multi-head self-attention operates over a single element and is identically 1, so MHSA cannot exchange information across time steps. The Time-Transformer is therefore a pointwise (per-sample) nonlinear map from the fused CLIP embedding to the timestamp. The claim that it 'captures implicit temporal progression' or 'learns temporal dependencies' is, by the paper's own equations, a renaming of static regression: there is no temporal sequence, no positional encoding, and no cross-time attention. The temporal language adds no content beyond the static feature-to-time mapping.

1 more flagged steps
  1. self citation load bearing [Section 2.1, Section 2.4, references [18], [28], [29]]
    "images are first encoded using CLIP's vision encoder, while corresponding textual prompts describing fungal stages are processed by the text encoder [18]. ... The dataset is analytically constructed to simulate clear distinctions between the developmental stages [29] ... made publicly available for research use [28]."

    The text prompts ([18]) and the dataset ([28], [29]) are prior work by the same authors, and no external biological image set or independently collected timestamp data is used anywhere in the evaluation. The central conclusion that CLIPTime 'effectively models biological progression' rests entirely on this self-citation chain: the generator defines the stages and timestamps, the authors' model is trained and tested on that same generated distribution, and the cited evidence for the dataset's realism is the authors' own construction. The load-bearing support is not an independent benchmark, so the biological-progression claim is self-referential.

full rationale

The paper's core claim of modeling biological progression is supported only by the authors' own synthetic fungal dataset, whose timestamps and stage boundaries are analytically constructed by the same team (refs [28], [29]). Training and evaluating on this dataset shows the model can fit the generator's explicit stage/timestamp coupling, but it does not demonstrate that real fungal growth is modeled; the regression target and the test set are both outputs of the same synthetic process. In addition, the Time-Transformer reshapes each sample to sequence length 1 and applies self-attention to a single token, where softmax is identically 1, so no temporal dependency across time steps is learned. The 'implicit temporal progression' language is therefore a renaming of a static pointwise regression. These two issues make the central biological-progression claim partially circular, while the classification accuracy on the synthetic test set remains internally consistent. Because no external benchmark is used and the central evidence chain reduces to self-generated data, the score is 6 rather than lower.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method rests on a self-generated synthetic benchmark, hand-set hyperparameters, and a regression head whose claimed temporal reasoning is not structurally supported. No external benchmark or real-world data is used.

free parameters (3)
  • min-max normalization bounds (t_min, t_max) = not reported numerically
    Used in Eq. 1 to scale timestamps to [0,1] and to re-project predictions; computed from the synthetic dataset, so the model's time scale is tied to the authors' generated data.
  • loss weights alpha, beta = 1.0
    Set by hand in Section 2.3 (Eq. 10); no ablation shows the balance is sensible.
  • Time-Transformer hidden dimension and layers = 2048 hidden, 2 layers
    Chosen without justification in Section 2.2; the regression head's capacity depends on these hand-set hyperparameters.
assumptions (4)
  • domain assumption CLIP's joint embedding space encodes semantically meaningful visual and textual features for fungal stage discrimination.
    The whole method relies on CLIP's pretrained representations being useful for fungal images, which the paper does not test.
  • ad hoc to paper The synthetic fungal dataset's timestamps and stage labels are accurate ground truth.
    Section 2.4 states the dataset is analytically constructed; the generator is described in prior self-cited papers [28], [29], and there is no external validation.
  • domain assumption Element-wise summation of the image and text embeddings preserves the information needed for both classification and regression.
    Section 2.2 uses a simple sum to fuse modalities without comparing to concatenation or cross-attention.
  • ad hoc to paper Self-attention over a sequence of length 1 can capture temporal dependencies.
    Eq. 2 fixes the sequence length to 1, so the multi-head self-attention layer has only one token and cannot pool information across time steps; the paper nonetheless claims it models temporal progression.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CLIPTime: Time-Aware Multimodal Representation Learning from Images and Text." pith.science (2026). https://pith.science/paper/XIZWKWMX

@misc{pith2026250800447,
  author       = {Pith},
  title        = {Pith review of: CLIPTime: Time-Aware Multimodal Representation Learning from Images and Text},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XIZWKWMX}},
  note         = {Machine review of arXiv:2508.00447}
}
read the original abstract

Understanding the temporal dynamics of biological growth is critical across diverse fields such as microbiology, agriculture, and biodegradation research. Although vision-language models like Contrastive Language Image Pretraining (CLIP) have shown strong capabilities in joint visual-textual reasoning, their effectiveness in capturing temporal progression remains limited. To address this, we propose CLIPTime, a multimodal, multitask framework designed to predict both the developmental stage and the corresponding timestamp of fungal growth from image and text inputs. Built upon the CLIP architecture, our model learns joint visual-textual embeddings and enables time-aware inference without requiring explicit temporal input during testing. To facilitate training and evaluation, we introduce a synthetic fungal growth dataset annotated with aligned timestamps and categorical stage labels. CLIPTime jointly performs classification and regression, predicting discrete growth stages alongside continuous timestamps. We also propose custom evaluation metrics, including temporal accuracy and regression error, to assess the precision of time-aware predictions. Experimental results demonstrate that CLIPTime effectively models biological progression and produces interpretable, temporally grounded outputs, highlighting the potential of vision-language models in real-world biological monitoring applications.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references · 20 canonical work pages

  1. [28]

    Synthetic Fungi Datasets: A Time-Aligned Approach

    Anju Rani, Daniel O Arroyo, and Petar Durdevic. Synthetic fungi datasets: A time-aligned approach. arXiv preprint arXiv:2501.02855, 2025

    work page Pith review arXiv 2025

  2. [29]

    Fungalzsl: A fine-grained synthetic dataset for zero-shot fungi classification

    Anju Rani, Daniel O Arroyo, and Petar Durdevic. Fungalzsl: A fine-grained synthetic dataset for zero-shot fungi classification. https://data.mendeley.com/datasets/rw6ndgyrd7/1, 2025. Accessed: 2025-02-18. 11

    work page 2025

  3. [1]

    Large language models: A survey

    Shervin Minaee, Tomas Mikolov, Narjes Nikzad, Meysam Chenaghlu, Richard Socher, Xavier Amatriain, and Jianfeng Gao. Large language models: A survey. arXiv preprint arXiv:2402.06196, 2024

    arXiv 2024

  4. [2]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pages 8748–8763. PMLR, 2021

    2021

  5. [3]

    Vlm agents generate their own memories: Distilling experience into embodied programs of thought

    Gabriel Sarch, Lawrence Jang, Michael Tarr, William W Cohen, Kenneth Marino, and Katerina Fragkiadaki. Vlm agents generate their own memories: Distilling experience into embodied programs of thought. Advances in Neural Information Processing Systems, 37:75942–75985, 2024

    work page 2024

  6. [4]

    Rethinking vlms and llms for image classification

    Avi Cooper, Keizo Kato, Chia-Hsien Shih, Hiroaki Yamane, Kasper Vinken, Kentaro Takemoto, Taro Sunagawa, Hao-Wei Yeh, Jin Yamanaka, Ian Mason, et al. Rethinking vlms and llms for image classification. Scientific Reports, 15(1):1–20, 2025

    work page 2025

  7. [5]

    Regression in eo: Are vlms up to the challenge? arXiv preprint arXiv:2502.14088, 2025

    Xizhe Xue and Xiao Xiang Zhu. Regression in eo: Are vlms up to the challenge? arXiv preprint arXiv:2502.14088, 2025

    arXiv 2025

  8. [6]

    From images to signals: Are large vision models useful for time series analysis? arXiv preprint arXiv:2505.24030, 2025

    Ziming Zhao, ChengAo Shen, Hanghang Tong, Dongjin Song, Zhigang Deng, Qingsong Wen, and Jingchao Ni. From images to signals: Are large vision models useful for time series analysis? arXiv preprint arXiv:2505.24030, 2025

    arXiv 2025

Show all 29 references
  1. [7]

    Multi-step time series forecasting with an ensemble of varied length mixture models

    Yicun Ouyang and Hujun Yin. Multi-step time series forecasting with an ensemble of varied length mixture models. International journal of neural systems, 28(04):1750053, 2018

    2018

  2. [8]

    Cogvlm2: Visual language models for image and video understanding.arXiv preprint arXiv:2408.16500, 2024

    Wenyi Hong, Weihan Wang, Ming Ding, Wenmeng Yu, Qingsong Lv, Yan Wang, Yean Cheng, Shiyu Huang, Junhui Ji, Zhao Xue, et al. Cogvlm2: Visual language models for image and video understanding.arXiv preprint arXiv:2408.16500, 2024

    2024 arXiv

  3. [9]

    Integrating llms with its: Recent advances, potentials, challenges, and future directions

    Doaa Mahmud, Hadeel Hajmohamed, Shamma Almentheri, Shamma Alqaydi, Lameya Aldhaheri, Ruhul Amin Khalil, and Nasir Saeed. Integrating llms with its: Recent advances, potentials, challenges, and future directions. IEEE Transactions on Intelligent Transportation Systems, 2025

    2025

  4. [10]

    itransformer: Inverted transformers are effective for time series forecasting

    Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. itransformer: Inverted transformers are effective for time series forecasting. arXiv preprint arXiv:2310.06625, 2023

    2023 arXiv

  5. [11]

    Pv-vlm: A multimodal vision-language approach incorporating sky images for intra-hour photovoltaic power forecasting

    Huapeng Lin and Miao Yu. Pv-vlm: A multimodal vision-language approach incorporating sky images for intra-hour photovoltaic power forecasting. arXiv preprint arXiv:2504.13624, 2025. 10 CLIPTime: Time-Aware Multimodal Representation Learning from Images and Text A PREPRINT

    2025 arXiv

  6. [12]

    Time-vlm: Exploring multi- modal vision-language models for augmented time series forecasting

    Siru Zhong, Weilin Ruan, Ming Jin, Huan Li, Qingsong Wen, and Yuxuan Liang. Time-vlm: Exploring multi- modal vision-language models for augmented time series forecasting. arXiv preprint arXiv:2502.04395, 2025

    2025 arXiv

  7. [13]

    Leveraging temporal contextualization for video action recognition

    Minji Kim, Dongyoon Han, Taekyung Kim, and Bohyung Han. Leveraging temporal contextualization for video action recognition. In European Conference on Computer Vision, pages 74–91. Springer, 2024

    2024

  8. [14]

    C-tpt: Calibrated test-time prompt tuning for vision-language models via text feature dispersion

    Hee Suk Yoon, Eunseop Yoon, Joshua Tian Jin Tee, Mark Hasegawa-Johnson, Yingzhen Li, and Chang D Yoo. C-tpt: Calibrated test-time prompt tuning for vision-language models via text feature dispersion. arXiv preprint arXiv:2403.14119, 2024

    2024 arXiv

  9. [15]

    Pattern recognition and prediction in time series data through retrieval- augmented techniques

    Nitya Singh and Aleena Swetapadma. Pattern recognition and prediction in time series data through retrieval- augmented techniques. In 2024 International Conference on Electrical Electronics and Computing Technologies (ICEECT), volume 1, pages 1–6. IEEE, 2024

    2024

  10. [16]

    Driver intention prediction using text prompts with in-cabin and out-cabin cameras

    Bin He, Ningmei Yu, and Hongcheng Han. Driver intention prediction using text prompts with in-cabin and out-cabin cameras. In 2024 International Conference on Cyber-Enabled Distributed Computing and Knowledge Discovery (CyberC), pages 40–43. IEEE, 2024

    2024

  11. [17]

    Diverse data augmentation with dif- fusions for effective test-time prompt tuning

    Chun-Mei Feng, Kai Yu, Yong Liu, Salman Khan, and Wangmeng Zuo. Diverse data augmentation with dif- fusions for effective test-time prompt tuning. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2704–2714, 2023

    2023

  12. [18]

    Fungalzsl: Zero-shot fungal classification with image caption- ing using a synthetic data approach

    Anju Rani, Daniel O Arroyo, and Petar Durdevic. Fungalzsl: Zero-shot fungal classification with image caption- ing using a synthetic data approach. arXiv preprint arXiv:2502.19038, 2025

    2025 arXiv

  13. [19]

    Fungal biology

    Jim W Deacon. Fungal biology. John Wiley & Sons, 2005

    2005

  14. [20]

    Assembling the fungal tree of life: progress, classification, and evolution of subcellular traits

    Franc ¸ois Lutzoni, Frank Kauff, Cymon J Cox, David McLaughlin, Gail Celio, Bryn Dentinger, Mahajabeen Padamsee, David Hibbett, Timothy Y James, Elisabeth Baloch, et al. Assembling the fungal tree of life: progress, classification, and evolution of subcellular traits. American...

    2004

  15. [21]

    Developments in fungal taxonomy

    Josep Guarro, Josepa Gen ´e, and Alberto M Stchigel. Developments in fungal taxonomy. Clinical microbiology reviews, 12(3):454–500, 1999

    1999

  16. [22]

    Assessment of fungal spores and spore-like diversity in environmental samples by targeted lysis

    Andrea Corona Ramirez, Dana ´e Bregnard, Thomas Junier, Guillaume Cailleau, Cristina Dorador, Saskia Bind- schedler, and Pilar Junier. Assessment of fungal spores and spore-like diversity in environmental samples by targeted lysis. BMC microbiology, 23(1):68, 2023

    2023

  17. [23]

    Architecture and developmental dynamics of the external mycelium of the arbuscular mycorrhizal fungus glomus intraradices grown under monoxenic conditions

    Berta Bago, Concepcion Azcon-Aguilar, and Yves Pich ´e. Architecture and developmental dynamics of the external mycelium of the arbuscular mycorrhizal fungus glomus intraradices grown under monoxenic conditions. Mycologia, 90(1):52–62, 1998

    1998

  18. [24]

    Study of kinetic model for fungal spore germination under dynamic conditions: Case study on germination of penicillium expansum spores

    Panuluk Boonruang and Nuttapol Lerkkasemsan. Study of kinetic model for fungal spore germination under dynamic conditions: Case study on germination of penicillium expansum spores. Heliyon, 9(12), 2023

    2023

  19. [25]

    Pathways of pathogenicity: transcriptional stages of germination in the fatal fungal pathogen rhizopus delemar

    Poppy CS Sephton-Clark, Jose F Mu ˜noz, Elizabeth R Ballou, Christina A Cuomo, and Kerstin V oelz. Pathways of pathogenicity: transcriptional stages of germination in the fatal fungal pathogen rhizopus delemar. Msphere, 3(5):10–1128, 2018

    2018

  20. [26]

    A model for growth of a single fungal hypha based on well-mixed tanks in series: simulation of nutrient and vesicle transport in aerial reproductive hyphae

    Wellington Balmant, Maura Harumi Sugai-Gu ´erios, Juliana Hey Coradin, Nadia Krieger, Agenor Furigo Junior, and David Alexander Mitchell. A model for growth of a single fungal hypha based on well-mixed tanks in series: simulation of nutrient and vesicle transport in aerial rep...

    2015

  21. [27]

    Neurospora crassa nadph oxidase nox-1 is localized in the vacuolar system and the plasma membrane

    Nallely Cano-Dom ´ınguez, Barry Bowman, Leonardo Peraza-Reyes, and Jes´us Aguirre. Neurospora crassa nadph oxidase nox-1 is localized in the vacuolar system and the plasma membrane. Frontiers in microbiology, page 1825, 2019

    2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.