Hoang, Phi-Hung, Tran, Van-Manh, Trinh, Nam-Thuan, Phan, Thi-Thu-Hong · Nov 6, 2026
Phi-Hung Hoang and coauthors present HMMF-Net for fish feeding assessment
Markdown · JSON · Original source
Phi-Hung Hoang and coauthors propose HMMF-Net, a lightweight system for automated fish feeding intensity assessment (FFIA), which they describe as important for precision aquaculture and feed cost optimization. They argue that video- or audio-only methods can be vulnerable to visual disturbances or acoustic noise, while some existing systems are too computationally intensive for edge devices. HMMF-Net combines multi-frame visual input with acoustic features, exchanging information through hierarchical bidirectional messenger cross-modal fusion (BMCF). On the AV-FFIA dataset, the authors report 96.77% accuracy and 96.88% F1-score on held-out test folds using five-fold cross-validation. They also report deployment on an Jetson Orin Nano, with an inference of 11.54 ms per 2-second sample. The paper presents combining sound and images as a way to assess feeding intensity on a resource-constrained device; its reported results concern this dataset and deployment setup.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
cs.AI · Article
Source details
Original source date: Nov 6, 2026 (2026-11-06).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-09T05:46:02.488Z. First observed: 2026-11-09T05:07:41.357Z. Retrieved: 2026-11-09T05:33:42.290Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The reported assessment results are from the AV-FFIA dataset and five-fold cross-validation; the evidence here does not establish performance on other datasets. The reported deployment result is specific to an Jetson Orin Nano using TensorRT with FP32 precision.
Open question
How does HMMF-Net perform on datasets or deployment conditions beyond the AV-FFIA test folds and Jetson Orin Nano setup?
Source-stated
The authors present HMMF-Net as a lightweight multimodal system for fish feeding intensity assessment, combining visual and acoustic input through BMCF.
we propose HMMF-Net, a lightweight multimodal architecture that combines early multi-frame visual encoding and multi-axis acoustic modeling through a hierarchical bidirectional messenger cross-modal fusion (BMCF) mechanism.
Source-stated
On the AV-FFIA dataset, the authors report 96.77% accuracy and 96.88% F1-score on held-out test folds using five-fold cross-validation.
Evaluated on the AV-FFIA dataset using five-fold cross-validation, HMMF-Net achieves 96.77% accuracy and 96.88% F1-score on the held-out test folds
Source-stated
The authors report deployment on an NVIDIA Jetson Orin Nano with 11.54 ms inference time per 2-second sample.
Deployment on an NVIDIA Jetson Orin Nano using TensorRT with FP32 precision maintains comparable accuracy, with an inference time of 11.54 ms per 2-second sample
Source-stated
The paper frames automated fish feeding intensity assessment as relevant to precision aquaculture and feed cost optimization.
Automated fish feeding intensity assessment (FFIA) is important for precision aquaculture and feed cost optimization.
Source and contributor context
Hoang, Phi-Hung, Tran, Van-Manh, Trinh, Nam-Thuan, Phan, Thi-Thu-Hong · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Multimodal model
A model that works with more than one kind of input or output, such as text, images and audio.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
NVIDIA
NVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: 1388e5aa-90eb-4994-8412-0dd4a68205b9.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-09T05:46:02.488Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: fadf7fd3881ec652b81abfafb67e80286d953dd1e9cfe51038beedc447443afd. Source text hash: 5bb6423639c7011fcbfffd775ffe866d13280f43591a17e786d4c3a76ec56c26.
Revision and source-check history
Version 1 · 2026-11-09T05:46:02.488Z
Phi-Hung Hoang and coauthors present HMMF-Net for fish feeding assessment
Phi-Hung Hoang and coauthors propose HMMF-Net, a lightweight system for automated fish feeding intensity assessment (FFIA), which they describe as important for precision aquaculture and feed cost optimization. They argue that video- or audio-only methods can be vulnerable to visual disturbances or acoustic noise, while some existing systems are too computationally intensive for edge devices. HMMF-Net combines multi-frame visual input with acoustic features, exchanging information through hierarchical bidirectional messenger cross-modal fusion (BMCF). On the AV-FFIA dataset, the authors report 96.77% accuracy and 96.88% F1-score on held-out test folds using five-fold cross-validation. They also report deployment on an Jetson Orin Nano, with an inference of 11.54 ms per 2-second sample. The paper presents combining sound and images as a way to assess feeding intensity on a resource-constrained device; its reported results concern this dataset and deployment setup.
Original source
cs.AI · Article
Source details
Original source date: Nov 6, 2026 (2026-11-06).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-09T05:46:02.488Z. First observed: 2026-11-09T05:07:41.357Z. Retrieved: 2026-11-09T05:33:42.290Z.
Limits
The reported assessment results are from the AV-FFIA dataset and five-fold cross-validation; the evidence here does not establish performance on other datasets. The reported deployment result is specific to an Jetson Orin Nano using TensorRT with FP32 precision.
Open question
How does HMMF-Net perform on datasets or deployment conditions beyond the AV-FFIA test folds and Jetson Orin Nano setup?
Source-stated
The authors present HMMF-Net as a lightweight multimodal system for fish feeding intensity assessment, combining visual and acoustic input through BMCF.
we propose HMMF-Net, a lightweight multimodal architecture that combines early multi-frame visual encoding and multi-axis acoustic modeling through a hierarchical bidirectional messenger cross-modal fusion (BMCF) mechanism.
Source-stated
On the AV-FFIA dataset, the authors report 96.77% accuracy and 96.88% F1-score on held-out test folds using five-fold cross-validation.
Evaluated on the AV-FFIA dataset using five-fold cross-validation, HMMF-Net achieves 96.77% accuracy and 96.88% F1-score on the held-out test folds
Source-stated
The authors report deployment on an NVIDIA Jetson Orin Nano with 11.54 ms inference time per 2-second sample.
Deployment on an NVIDIA Jetson Orin Nano using TensorRT with FP32 precision maintains comparable accuracy, with an inference time of 11.54 ms per 2-second sample
Source-stated
The paper frames automated fish feeding intensity assessment as relevant to precision aquaculture and feed cost optimization.
Automated fish feeding intensity assessment (FFIA) is important for precision aquaculture and feed cost optimization.
Source and contributor context
Hoang, Phi-Hung, Tran, Van-Manh, Trinh, Nam-Thuan, Phan, Thi-Thu-Hong · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Multimodal model
A model that works with more than one kind of input or output, such as text, images and audio.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
NVIDIA
NVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: 1388e5aa-90eb-4994-8412-0dd4a68205b9.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-09T05:46:02.488Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: fadf7fd3881ec652b81abfafb67e80286d953dd1e9cfe51038beedc447443afd. Source text hash: 5bb6423639c7011fcbfffd775ffe866d13280f43591a17e786d4c3a76ec56c26.
passed · 2026-11-09T05:46:02.488Z · anthropic/claude-opus-5.5
Publication: 1388e5aa-90eb-4994-8412-0dd4a68205b9
attribution — pass: The reading attributes the work to Phi-Hung Hoang and coauthors, matching the listed authors. The performance figures and the importance of fish feeding intensity assessment are kept as the authors' reported claims.
Authors:Phi-Hung Hoang, Van-Manh Tran, Nam-Thuan Trinh, Thi-Thu-Hong Phan
summary — pass: The title, summary and question match the abstract. The reading covers the motivation, the combination of video and audio, the reported results and the edge deployment. A general reader can follow the main point. The phrase 'resource-constrained device' conveys what an edge device is. Terms such as F1-score could use term-library tooltips, but they do not block understanding. The question raises a real open issue about generalization without loading it.
unimodal methods are vulnerable to acoustic noise and visual disturbances, while existing multimodal frameworks often rely on computationally intensive 3D convolutions or Transformers, limiting their use on edge devices
date — pass: The originalPublished date of 2026-11-06 matches the arXiv v1 submission date. The later feed date is not used.
[Submitted on 6 Nov 2026]
support — pass: Each claim is directly stated in the abstract, and the excerpts are exact. The limitations correctly confine the results to the AV-FFIA dataset and the Jetson Orin Nano, TensorRT and FP32 setup. The reading does not imply real-world benefits beyond what the source states.
Evaluated on the AV-FFIA dataset using five-fold cross-validation, HMMF-Net achieves 96.77% accuracy and 96.88% F1-score on the held-out test folds
media — pass: The selected media is the verified arXiv abstract page for this paper. It is the direct original source and is supplied among the candidates.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
A model that works with more than one kind of input or output, such as text, images and audio.
Definition referenceNVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
Explore NVIDIATIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivNVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
Explore NVIDIANVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
Explore NVIDIAA model that works with more than one kind of input or output, such as text, images and audio.
Definition referenceNVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
Explore NVIDIATIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivNVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
Explore NVIDIANVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
Explore NVIDIA