---
title: PlantCLEF 2025 Challenge
url: https://www.emergentmind.com/topics/plantclef-2025-challenge
type: topic
---

# PlantCLEF 2025 Challenge

PlantCLEF 2025 was the LifeCLEF Lab at CLEF 2025 challenge on **multi-species plant identification in vegetation quadrat images**, formulated as a **weakly labelled multi-label classification problem** in which the goal was to predict all species present in a quadrat image using mainly single-label training data [2509.17602]. The benchmark combined a new test set of **2,105 high-resolution multi-label images annotated by experts** with a large training set of **1.4 million individual plant images**, and it distributed organizer-provided vision transformer models pre-trained on this data [2509.17602]. The challenge extended the multi-species vegetation-plot setting introduced in PlantCLEF 2024 and raised the best private score from **0.2873** in 2024 to **0.3648** in 2025, while still leaving the task far from solved [2509.17602][2509.15768].

## 1. Ecological setting and challenge definition

PlantCLEF 2025 was motivated by a practical ecological bottleneck. Quadrat images are essential for ecological studies because they enable **standardized sampling**, **the assessment of plant biodiversity**, **long-term monitoring**, and **large-scale field campaigns** [2509.17602]. In routine vegetation monitoring, botanists examine standardized plots—typically around **50 × 50 cm**, and in the overview also described as plots of **fifty centimetres or one square meter in size**—and identify all species present [2509.17602]. The challenge asked whether AI could identify all plant species visible in a single high-resolution quadrat image rather than requiring specimens to be photographed one by one [2509.17602].

This setting was technically distinctive within PlantCLEF. Earlier global-scale editions such as PlantCLEF 2022 and 2023 were observation-level species retrieval problems over **80,000 plant species**, whereas PlantCLEF 2024 and 2025 focused on **community-level species presence** in vegetation plots [2509.17622][2509.17602]. The 2025 task therefore combined a **label structure mismatch**—training data are largely **single-label**, test data require **multi-label** output—with a **domain gap** between close-up individual-plant images and vertical or near-vertical quadrat scenes [2509.17602].

The ecological scenes themselves were difficult. Quadrat images were described as top-view, high-resolution, multi-species scenes with overlapping plants, variable density, clutter, partial visibility, shadows, blur, and strong differences in phenological stage [2509.17602]. The overview reports **about 8 species per quadrat on average**, a density high enough that a score around **0.36** corresponded roughly to **4–5 correct predictions per image with relatively few false positives** [2509.17602].

## 2. Data resources and supervision mismatch

The main training resource was a large subset of **Pl@ntNet** data focused on **South-Western Europe** [2509.17602]. Its reported statistics were:

- **1,408,033 images**
- **1,151,904 observations**
- **7,806 species**
- **1,446 genera**
- **181 families** [2509.17602]

The split provided for single-plant model development contained **1,308,899** training images, **51,194** validation images, and **47,940** test images, with splitting performed at the **observation level** so that all images from a single observation belonged to the same split [2509.17602]. These splits were explicitly distinct from the hidden challenge test set of quadrat images [2509.17602].

The challenge test data consisted of **2,105 high-resolution quadrat images** compiled from several expert-produced datasets in **Pyrenean, Mediterranean, and South-Western European temperate floras** [2509.17602]. The official overview describes the test set as covering **around 400 species** [2509.17602]. Participant papers, however, characterize the effective test flora as **over 800 species** or **roughly 800 species** [2507.06093][2508.10457]. This suggests differing counting conventions across official and participant descriptions.

Acquisition conditions varied substantially. The quadrats could include **wooden frames** or **measuring tape**, could be more or less perpendicular to the ground depending on slope, and could contain shadows, blur, and weather-related quality changes [2509.17602]. These properties mattered because the dominant training imagery emphasized individual plants or plant organs, whereas the test imagery emphasized mixed-species vegetation patches [2509.17602].

To support domain adaptation, organizers also provided an unlabeled pseudo-quadrat resource: **212,782 images** from the **LUCAS Cover Photos 2006–2018 archive**, curated as high-resolution vegetation photos framed similarly to quadrats but not strictly limited to a **50 × 50 cm** area [2509.17602]. The overview notes that this complementary pseudo-quadrat dataset was **not used by any of the teams that submitted working notes** [2509.17602].

## 3. Evaluation methodology and official challenge resources

The official ranking criterion was an **F1 score** with **transect-aware averaging** [2509.17602]. Per-image F1 was defined as

$$
F1_i = \frac{2 \cdot \text{Precision}_i \cdot \text{Recall}_i}{\text{Precision}_i + \text{Recall}_i},
$$

with

$$
\text{Precision}_i = \frac{\text{TP}_i}{\text{TP}_i + \text{FP}_i}, \qquad
\text{Recall}_i = \frac{\text{TP}_i}{\text{TP}_i + \text{FN}_i}.
$$

The final challenge score averaged first within transects and then across transects:

$$
F1_{\text{avg-transect}} = \frac{1}{T} \sum_{k=1}^{T} \left( \frac{1}{Q_k} \sum_{j=1}^{Q_k} F1^{(k)}_j \right)
$$

where \(T\) is the number of transects and \(Q_k\) the number of quadrat images in transect \(k\) [2509.17602]. This choice was ecologically motivated because it prevented heavily sampled transects from dominating the final score [2509.17602].

The challenge was hosted on **Kaggle** [2509.17602]. Metadata use and external data were allowed, but each such run had to be paired with an equivalent run using only the provided visual data, so that the contribution of metadata or external resources could be measured fairly [2509.17602].

A central official resource was the release of two organizer-provided **Vision Transformer** models, both based on **ViT base patch-14 with registers**, initialized from **DINOv2** self-supervised pretraining on **LVD-142M** and then fine-tuned on the PlantCLEF individual-plant training data using **timm** [2509.17602]. These were:

- **ViTD2PC24OC** (`vit_base_patch14_reg4_dinov2_lvd142m_onlyclassifier`), in which only the classification head was trained.
- **ViTD2PC24All** (`vit_base_patch14_reg4_dinov2_lvd142m_onlyclassifier_then_all`), initialized from the previous model and then fine-tuned on all layers [2509.17602].

On the held-out single-plant test split, **ViTD2PC24OC** reached **Top-1 = 63.69** and **Top-5 = 83.88**, while **ViTD2PC24All** reached **Top-1 = 75.91** and **Top-5 = 92.82** [2509.17602]. The official **tiling-based baseline** achieved a macro-averaged F1 of **0.21708** on the challenge benchmark [2509.17602].

## 4. Dominant methodological pattern: tiling, aggregation, and inference-time adaptation

The clearest 2025 pattern was the dominance of **tile-based inference**. The official overview states that **all but two teams** cut each high-resolution quadrat image into smaller tiles, ran a classifier on each tile, and then aggregated predictions across tiles [2509.17602]. This was a direct response to the supervision mismatch: a model trained on images of individual plants was more effective when parts of a quadrat were made to resemble the training domain [2509.17602].

The second-place DS@GT system made this logic explicit. It combined **ViTD2PC24All**, **\(4 \times 4\) tiling**, tile-level classification, **visual-cluster priors** from **PaCMAP + K-Means**, and **geolocation filtering**, all without additional training [2507.06093]. A typical quadrat was about **2000 px** wide, so a **\(4 \times 4\)** partition produced tiles of roughly **500 px** per side, closely matching the model’s **\(518 \times 518\)** expected input size [2507.06093]. In the DS@GT ablation, full-image inference with a plain ViT gave **Private 0.00633** and **Public 0.01157**, whereas **ViT, top-9, 4x4** gave **Private 0.34420** and **Public 0.30810** [2507.06093]. This suggests that scale matching between tile size and receptive field was one of the central 2025 design principles.

The third-place Chlorophyll Crew system also treated multi-label prediction as an inference-time construction over local evidence rather than as end-to-end scene parsing. It used a frozen **DINOv2 ViT-B/14** backbone with multiple heads for **species**, **genus**, and **family**, combined with **multi-scale non-overlapping tiling**, border cropping to remove non-plant artifacts, one-species-per-tile constraints, and **dynamic threshold optimization based on mean prediction length** [2508.10457]. Their best-performing submissions used scales **4 and 5**, meaning **\(4 \times 4\)** and **\(5 \times 5\)** tilings, and dynamically targeted a mean prediction length around **4 species per image** [2508.10457].

A more divergent fifth-place strategy came from ADAM, which replaced standard tiling by **prototype-guided zero-shot segmentation** [2512.19957]. The method extracted **DINOv2 embeddings** from the training set, formed **\(K = 7806\)** K-Means centroids as prototype targets, trained a narrow ViT on test-image patches to reconstruct a **\(7806 \times 768\)** prototype matrix, and then used attention maps as a proxy localization mechanism before classification [2512.19957]. Its central empirical lesson was that **context matters much more than isolated patch classification**: direct patch-wise classification was weak, while **\(K \times K\)** contextual grid assembly around high-attention patches was much stronger [2512.19957].

## 5. Leaderboard, participation, and representative systems

PlantCLEF 2025 was organized on Kaggle and drew **540 initial entrants**, **55 participants** grouped into **38 teams**, **659 runs**, and **10 teams** that submitted working notes [2509.17602]. The top of the challenge was methodologically concentrated but not uniform.

| Rank | Team | Characteristic strategy |
|---|---|---|
| 1 | TheHeartOfNoise | Image preprocessing, JPEG recompression, quality and YCbCr subsampling, tiling |
| 2 | DS@GT-LifeCLEF | \(4 \times 4\) tiling, species-frequency aggregation, geolocation filtering, region-specific Bayesian reweighting |
| 3 | Chlorophyll Crew | Multi-head species/genus/family ViT, multi-scale tiling, dynamic thresholding |
| 4 | webmaking | Multi-scale tiling, GroundingDINO + SAM, seasonal filtering, ecological adjustment, cross-year aggregation |
| 5 | ADAM | Prototype-guided zero-shot segmentation and contextual patch-grid classification |

The winning system, **TheHeartOfNoise**, was notable because it pushed low-level image preprocessing unusually far. According to the overview, the participant focused heavily on **JPEG recompression** and the tuning of **quality** and **YCbCr subsampling**, and these choices substantially improved F1 by reducing the train-test distribution gap and acting as a form of regularization [2509.17602]. The official overview reports a best private leaderboard score of **0.3648**, and also states that the highest purely visual performance, **0.35016**, was obtained by TheHeartOfNoise using **test-time image preprocessing** and **multi-scale tiling with nine different sizes** [2509.17602].

The **DS@GT** system achieved **second place** with a best private macro-F1 of **0.34834** [2507.06093]. Its best variant, **ViT + PRIORS**, re-weighted tile predictions using cluster-specific priors estimated by averaging model posterior vectors over **3** K-Means clusters formed from **[CLS]** embeddings of the test images [2507.06093]. A related geolocation filter reduced the candidate label set from **7,806** to **4,981** species by retaining only species whose nearest geotagged observation was within **France, Spain, Italy, or Switzerland** [2507.06093].

**Chlorophyll Crew**, the **third-place** team, reported **2nd place on the public leaderboard** and **3rd place on the private leaderboard** [2508.10457]. Their best private score among listed submissions was **0.34575** for a **5h1l** model with target mean length **4.0**, max length **\(\infty\)**, tiling scales **4, 5**, and crop **10%**, while the best private score among their five selected final submissions was **0.33655** [2508.10457]. The system’s main contributions were taxonomy-aware multi-head prediction and explicit calibration of prediction set size through mean-length-controlled thresholding [2508.10457].

The **webmaking** system, ranked **4th**, combined **multi-scale tiling**, **test-time augmentation**, **GroundingDINO + SAM** to down-weight non-vegetal regions, **seasonal filtering** using **GBIF occurrence data**, ecological adjustment using **Ellenberg Indicator Values for Europe (EIVE)**, and **cross-year aggregation** [2509.17602]. It was one of the strongest examples of ecological prior integration in the challenge [2509.17602].

The **ADAM** system, ranked **5th**, achieved **\(F_1 = 0.33331\)** on the private leaderboard [2512.19957]. Its prototype-guided attention model showed that attention-based localization could be competitive, but only when classification restored enough local context through **\(K = 9\)** contextual patch grids [2512.19957].

## 6. Recurring themes, misconceptions, and unresolved problems

Several methodological conclusions recur across the official overview and the working notes. The first is that **PlantCLEF 2025 was not mainly a problem of inventing a new classifier, but of adapting an existing plant recognizer to a radically different test domain** [2507.06093]. Tiling, aggregation, cropping, priors, and low-level preprocessing repeatedly mattered more than architectural novelty alone [2509.17602][2508.10457].

A common misconception was that the problem could be addressed by ordinary whole-image plant classification. Participant ablations strongly contradicted that view. In DS@GT, full-image inference was almost useless, whereas quadrat tiling produced the decisive performance jump [2507.06093]. The overview likewise reports that most successful teams converged on some form of tiled high-resolution inference [2509.17602].

A second misconception concerned “zero-shot” claims. Both DS@GT and ADAM described their methods as **training-free** or **zero-shot** relative to the **2025 task**, but both relied on organizer-provided **DINOv2-based ViT models** already fine-tuned on PlantCLEF data from earlier training resources [2507.06093][2512.19957]. In other words, the systems were zero-additional-training for the quadrat task, not zero-shot in the absolute sense.

The challenge also exposed limits of segmentation and detection. The overview concludes that **vegetation-focused filtering did not consistently yield strong measurable gains**, and that conventional object detection was poorly matched to dense vegetation because quadrat scenes have **no clear object boundaries**, **strong overlap**, **many indistinct plant instances**, and **very fine-grained species differences** [2509.17602]. This was one reason why systems based on detection-style localization were not dominant, despite the apparent naturalness of localizing plants in cluttered scenes [2509.17602].

Public/private divergence was another central issue. The official overview reports large leaderboard shifts: **TheHeartOfNoise** moved from **3rd public** to **1st private**, **webmaking** fell from **1st public** to **4th private**, and **DS@GT PlantCLEF** rose from **9th public** to **2nd private** [2509.17602]. This established leaderboard overfitting as a real concern and partly explains why many high-scoring systems emphasized robust priors and conservative aggregation rather than purely public-score-driven tuning.

Substantial headroom remained. The overview identifies open directions such as **multi-label models that process full-resolution quadrat images end-to-end**, robustness to **low-level encoding and compression differences**, stronger integration of **ecological priors** and **structural priors**, and better **domain adaptation** between individual-plant and quadrat-image domains [2509.17602]. Participant papers added more specific proposals, including **sliding-window inference**, **adaptive receptive fields**, **token merging**, **self-training on pseudo-labels**, and **CNN/ViT ensembling** [2507.06093]. PlantCLEF 2025 therefore marked both a major empirical advance and a clear statement that weakly supervised plant community recognition in quadrat imagery remained an open research problem.

Source: https://www.emergentmind.com/topics/plantclef-2025-challenge