. . . . "2026-04-22"^^ . . "Decrop 2025 phytoplankton CNN metrics reproduced within 0.003 percentage points" . "The independent reproduction confirms the headline classification metrics reported by Decrop et al. 2025. All five reported metrics — top-1 accuracy, top-5 accuracy, micro F1, macro F1, and weighted F1 — match the paper's reported values within 0.003 percentage points. The Decrop 2025 phytoplankton CNN classifier is reproducible end-to-end from the publicly released artefacts (pretrained model, dataset, code package), with no deviations of substance from the original methodology." . . "Reproduction performed on the 33,718 images of the released test partition using the authors' planktonclas code package and the publicly released EfficientNetV2-B0 weights. Computed metrics versus paper-reported values:\n \n - Top-1 accuracy: 86.3426% (paper: 86.34%; delta +0.003 pp)\n- Top-5 accuracy: 98.7633% (paper: 98.76%; delta +0.003 pp) \n- Micro F1: 86.3426% (paper: 86.34%; delta +0.003 pp)\n- Macro F1: 78.7583% (paper: 78.76%; delta −0.002 pp) \n- Weighted F1: 86.2470% (paper: 86.25%; delta −0.003 pp)\n \nAll five deltas are below 0.005 percentage points — within the precision the paper itself reports. The full results JSON and predicted probabilities matrix are archived at Zenodo 10.5281/zenodo.19701133. Inference time was approximately 27 minutes for 33,718 images on CPU (Apple M1 Pro). Github repository: https://github.com/annefou/fiesta-decrop-reproduction" . "This study reproduces test-set evaluation only; it does not retrain the model. It therefore does not address whether the training procedure itself is reproducible — a future replication that retrains EfficientNetV2-B0 from scratch using the released code and splits would address that. The study also evaluates only on the released test.txt; the model's behaviour on truly out-of-distribution data (other water bodies, other plankton imaging instruments, other taxonomic mixes) is not assessed." . . . . . "Anne Fouilloux" . "2026-04-26T13:58:19.370Z"^^ . . . . . "NP created using Declaring a replication study outcome according to FORRT" . . "RSA" . "MIICIjANBgkqhkiG9w0BAQEFAAOCAg8AMIICCgKCAgEAoDcOiD+jen8awiJ6DB2ewDw66PeG64hODmgNFwy7GrwQui4HKnHdvxd++1UhTgiOfycxyxBb7sXPSikLw/1TsSyPsEl0P3/+600szxpTGgLNzW+bZ2DVP3d8ERMV1aWpH0ci3B/5vmK+vXQZ4uCoq57NE0MiFg5c13Gy0gd6n7wZYEhYM4AjWSLL0QS/HY+TFZMYL9bCFeATennGrlB2UEjRlw21UB2Ah16ZZ6hxQlfctFJZE7TGnBJPB3ttTjfcOfamhjZVwQ0yV9mv7x6PGiSmkzpJTVLjn8hagoKT05YUwVQArFb+w7f6sXqvvljMigjd/Rbqgbye/lLUAZLfJSnFM58TubfpEJvXV4zNMDEoT3VQ7dokgoLgMrmjZCKATtQ7gomocoTJ1NhN2esRNtGzWaS2obL/mueUQlMlavssZnqL8WICkdAuDlwDVNbsbwEWKQ50kiPdAdduSigifxA4CM7TgvnxqZVoAResEGP6UhTTem3T4CsbEas1Caj9wa7M1jPjACu5LF5BwcVns3ZQHWLipjRjD+9/ur3G8QtuxbNhmXlDYQ6tXxB1lK+Oz7O519b3bA15ilzFl0SdvMBGTe46xaQ9DsJT18THKnPbUhNMy0dH0VtzpB+EEaXZ25Fp9VHMEUqo1lLS9e89eO3efiqkESKQ7wmB+/DlIRcCAwEAAQ==" . "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" . . .