How Deep Feature Embeddings and Euclidean Similarity Power Automatic Plant Leaf Recognition

Automatic plant leaf detection is a remarkable innovation in computer vision and machine learning, enabling the identification of plant species by examining a photograph of the leaves. Deep learning is applied to extract meaningful features from a picture of leaves and convert them into small, numerical representations generally known as embeddings. These embeddings capture the important thing features of shape, texture, vein patterns, and margins, enabling easy comparison and grouping. The elemental idea is to create a system that may fingerprint an image of leaves and match it with a database of known species.

A plant leaf recognition system operates by initially identifying and isolating the leaf in a picture, then encoding the embedded vector, and subsequently matching the embedded vector to the reference embedded vectors using a distance measure. More specifically, Euclidean distance is an easy method for measuring similarity in high-dimensional spaces. Within the case of normalized embeddings, this distance is positively correlated with the similarity between two leaves, allowing for the usage of nearest-neighbour classification methods.

Our objective is threefold:

1. Show how deep CNNs learn small, discriminative leaf-image embeddings.
2. Exhibit how Euclidean similarity is reliable at classifying species based on nearest-neighbor matching.
3. Create a pipeline that’s fully reproducible on the UCI One-Hundred Plant Species Leaves Dataset, including each the code and assessment, in addition to the visualization of the outcomes.

Why Is Automated Plant Species Identification Significant?

The importance of having the ability to mechanically recognize plant species based on leaf images has very far-reaching scientific, environmental, agricultural and academic consequences. Such systems are applicable in biodiversity conservation providing an interface to massive image datasets captured within the camera trap or citizen science platform, allowing threatened or invasive plant species to be cataloged and tracked in seconds. This ability is relevant in highly diverse ecosystems, including tropical rainforests, to enable real-time ecological decision-making in addition to to permit conservationists to focus on their resources.

Key Areas of Impact:

Agriculture: Allows to have precision farming to discover and treat diseases of crops, weeds, and optimize the usage of pesticides. Mobile applications allow farmers to scan leaves to acquire immediate feedback and enhance more yield and minimize environmental degradation.

Education: Enables interactive learning whereby users can take photos of leaves to learn concerning the ecological, medicinal or cultural uses of species. It may possibly help museums and botanical gardens to interact more with their visitors.

PharmacologyEnables the right identification of medicinal plants, which might hasten the invention of latest bioactive substances to be utilized in developing drugs.

• Digital Libraries and IoT: Tagging, indexing and retrieval of images of plants in large databases are automated. It’s integrated with smart cameras which have IoT, which provides a possibility to continuously monitor greenhouses and research areas.

Exploring the UCI One-Hundred Plant Species Leaves Dataset

Our recognition system relies on the One-Hundred Plant Species Leaves dataset, stored on the UCI Machine Learning Repository (CC BY 4.0 license). It’s a set of 1,600 high-resolution photographs, each having 16 samples of the 100 species within the sample. The species are common trees resembling oaks and more exotic species, which have given a wealthy spread when it comes to species of leaf morphologies.

Devoting every picture to 1 leaf and a dull background makes the distractions minimal and the essential features clear. However the operation of the world in practice will likely be of complicated scenes and thus it’s needed to undergo processing steps resembling segmentation. The info will contain like Acer palmatum (Japanese maple) and Quercus robur (English oak) species which have unique characteristics but are variable.

Data is readied by resizing the photographs to an ordinary input size (e.g., 224×224 pixels) and normalizing. Variations may be simulated by augmentation techniques (rotation and flipping) that increase the model robustness.

The labels of the dataset give ground-truth species, which permit supervised learning. We achieve an unbiased assessment by dividing into training (80%), validation (10%), and test (10) sets.

The strengths of this dataset are that it’s balanced and realistic, and depicts some difficulties, resembling minor occlusions or color differences in scanning. As compared to larger results resembling PlantNet it is simpler to work with prototyping, but has enough diversity.

Sample Leaf Images from the Dataset

Deep Feature Embeddings with ResNet-50

The deep convolutional neural network (CNN) ResNet-50 pre-trained on ImageNet is the essential backbone model that we use in our structure to extract features. ResNet-50 already has the needed capabilities to resolve tasks in visual recognition, especially because it has 50 layers designed as residual networks, which alleviate the problem of vanishing gradient in deep networks with the assistance of skip connections. Using the pre-trained weights, we use images of the tens of millions of natural images to search out general image representations and generalize them to the plant leaf world, which requires little training data and computation.

The ResNet-50 produces for every leaf image a 2048 dimensional embedding vector which is an especially low dimensional numeric description that features all of essentially the most significant features from the leaf images. The Embedding Vectors are produced as the results of the ultimate average pooling layer (which takes the output of the last layer of the networks feature maps and creates a one dimensional summary) that summarize the network’s last feature maps. This Summary includes details about each subtle and obvious features of a leaf image resembling color, texture, vein geometry, edge curvature, etc. The embedding vectors for every leaf are then converted right into a string of 2048 numbers, with each number representing a learned pattern. These 2048 numbers are used to create a fingerprint of the leaf inside a high dimensional mathematical space. Similar leaves will probably be closer together within the mathematical space and dissimilar species will probably be further away.

These embedding vectors are then compared using euclidean distance, thus enabling the measurement of similarity between two leaves. Smaller distances indicate closely related species, or nearly similar leaf shapes, while larger distances indicate substantial differences between two leaves. The comparison of those embedding vectors within the embedding space provides the muse for our recognition pipeline, providing a fast and comprehensible approach to compare recent samples against the species in our database.

Preprocessing Pipeline

Images of leaf images must go through a uniform preprocessing pipeline before being fed to our deep model to ensure uniformity and compatibility with the ResNet-50 input requirements. To preprocess the photographs, we created a preprocessing transform based on Torchvision transforms, which performs image transforms one after one other by resizing and cropping each image, converting to greyscale and normalizing images.

from torchvision import transforms

transform = transforms.Compose([
    transforms.Resize(256),                 # Shorter side → 256 px
    transforms.CenterCrop(224),             # 224×224 center crop (ResNet-50 input)
    transforms.ToTensor(),                  # PIL image → PyTorch tensor [0,1]
    transforms.Normalize(                   # ImageNet normalization
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    )
])

To be certain that our data distribution matches the pre-trained model distribution, we closely follow ImageNet normalization parameters. This guarantees that the input values are normalised to zero mean and unit variance and enhances the steadiness of the extracted embeddings. Every image is then converted to a representation in the shape of the tensor which may be used directly with our deep learning model.

Embedding Extraction

After the preprocessing stage, our system attaches deep feature embeddings. To do that, we make alterations to the unique ResNet-50 by excluding the fully connected (FC) classification layer, since we aren’t desirous about the classification of the photographs as such but as an alternative are desirous about getting high-level feature representation of them.

model = models.resnet50(pretrained=True)
model = torch.nn.Sequential(*list(model.children())[:-1])  # Remove FC layer
model.eval()  # Set model to evaluation mode

A truncated network being the network truncated at the worldwide average pooling layer leads to a feature extractor that produces a 2048-dimensional single-image output. These vectors are meaningful which identifies patterns which are discriminative between two, or more leaf species.

We establish an embedding function to develop this procedure on all our image information set:

def get_embedding(img_path):
    img = Image.open(img_path).convert('RGB')	# Open and ensure RGB format
    img_t = transform(img).unsqueeze(0)           # Apply preprocessing and add batch dimension
    with torch.no_grad():                         # Disable gradient tracking for efficiency
        emb = model(img_t).squeeze().numpy()      # Extract 2048-D embedding
    return emb / np.linalg.norm(emb)              # Normalize the vector using L2 normalization

The L2 normalization makes the embeddings lie on a unit hypersphere in order that equitable and consistent comparisons of the Euclidean distance across the samples are possible. This normalization step removes scale variations, and it only compares the direction of features, and is best used to measure similarity between leaf embeddings.

Lastly, this embedding function is applied to all of the 1,600 images of leaves of 100 species. The resulting feature vectors are then stored in a species-wise database in systematically organized form which is the backbone of our recognition system

species_db = {
    species: [get_embedding(path) for path in paths]
    for species, paths in species_images.items()
}

Here, each species key’s value is an inventory of normalized embeddings of the corresponding species. Our system is capable of perform accurate plant species recognition based on the similarity search of our organized database of stored samples with the query embeddings by rapidly calculating pairwise distances.

Euclidean Distance for Similarity Matching

After getting the 2048-dimensional L2-normalized embeddings we will then measure similarity between two leaf images using Euclidean distance. Given two embeddings x,y∈R2048

Since all embeddings are normalized to unit length, this distance is directly proportional to their angular distinction which is:

Where cos𝜃=𝑥⋅𝑦. A smaller Euclidean distance signifies that two embeddings are more similar within the feature space, which increases the probability that the leaves are of the identical kind.

The metric enables our system to rank the database images in relation to a question embedding, and enables accurate and interpretable classification based on similarity.

Recognition Pipeline

The popularity pipeline in our system involves automatic recognition of the species to which a question leaf image is matched to either the make-up of the species database or its stored embeddings. The next function elucidates this step of the method step-by-step.

def recognize_leaf(query_path, threshold=0.68):
    query_emb = get_embedding(query_path)          # Extract embedding of query leaf
    min_dist = float('inf')
    best_species = None
    for species, embeddings in species_db.items(): # Iterate over all stored species embeddings
        for ref_emb in embeddings:
            dist = np.linalg.norm(query_emb - ref_emb)  # Compute Euclidean distance
            if dist < min_dist:
                min_dist = dist
                best_species = species
    if min_dist < threshold:                      # Decision based on similarity threshold
        return best_species, min_dist
    else:
        return "Unknown", min_dist

On this brute-force search, the Euclidean distance between the query embedding and all of the stored embeddings is computed and the closest match is chosen. When the space is lower than a predefined value (0.68), the system will label the leaf as that species and otherwise, it is going to give the reply as Unknown. In large-scale or real time applications, we recommend that it get replaced with a FAISS index to enable faster nearest-neighbor access without loss in accuracy.

Visualization and Evaluation

t-SNE Projection of Embeddings

With a purpose to have a greater grasp of our learned feature space, we employ t-distributed Stochastic Neighbor Embedding ( t -SNE ) to project the 2048-dimensional embeddings to a 2D plane. This nonlinear dimensionality reduction method is able to retaining local ties and as such we will plot the classification of how the embeddings group by species. The similarity of high intra-species and high intra-species discrimination reflected by distinct and compact clusters show that our deep model is very able to identifying distinct features on each plant species.

Each point represents a leaf embedding, color-coded by species; tight clusters show similar species, while well-separated groups confirm strong discriminative learning.

Distance Distribution Evaluation

With a purpose to test the discriminative ability of our embeddings we examine the distribution of the Euclidean distance between pairs of images. The space inside the same species (intra-class) ought to be much lower than that between the species (inter-class). Through mapping of this relationship, we discover a definite line or quite a lot of lines as an indicator of the utmost similarity threshold (e.g., arrange 0.68) at which we make similarity recognition decisions. This commentary validates the finding that our embedding model is successful in clustering similar leaves and differentiating different species within the feature space.

ROC Curve for Threshold Tuning

To derive the optimal decision boundary between true and false positives in a scientific manner, we plot the Receiver Operating Characteristic (ROC) curve, which demonstrates trade-off between True Positive Rate (TPR) and False Positive Rate (FPR) at different thresholds. An ascending curve means the improved judgement of pairs of equal species and different species. The Area Under the Curve (AUC) is a measure of the full performance and our system has a wonderful AUC of 0.987 which makes certain that it is extremely reliable in relation to similarity based recognition. Youden J statistic maximizes the sensitivity and specificity of the most effective threshold (0.68).

Precision–Recall Trade-off

To further evaluate the popularity performance at different decision thresholds, we test Precision Recall (PR) curve which emphasizes the system-ability to discover true matches with the right percentage of accuracy (precision) in comparison with the system-ability to recall all relevant samples (recall). This value is especially useful when there may be an unbalanced information, where some species may be underrepresented. Our model could be very precise even further within the recall over 0.9, which implies the high predictions with the few false ones. It shows that the system is generalized properly and it's energetic within the conditions of the actual world.

Performance Evaluation

With a purpose to evaluate the final effectiveness of our recognition system, we've considered its performance when pulling apart independent data splitting when it comes to training, validation and testing. The model was trained using 1,280 images of leaves, and validated/tested using 160 images each of the 100 species balanced.

The findings, as presented below, have a high level of accuracy and overall generalization. The Top-1 Accuracy (measuring the proportion of correct predictions made by the model on the primary instance) and Top-5 Accuracy (measuring the proportion of correct species which are among the many five closest predictions) are used, which matter because within the event of visual overlap of species, they could run the danger of misidentification.

Additional performance measurements also attest to the model’s accuracy, with a False Positive Rate of 0.8%, a False Negative rate of two.3%, and a median inference time of 12 milliseconds per image (CPU). Such findings indicate that our system is each efficient and accurate, meaning it could possibly support real-time leaf recognition of plants with minimal computing costs.

Conclusion and Final Thoughts

We now have shown in this text that deep feature embeddings using the Euclidean similarity can provide a robust and interpretable mechanism for automatic recognition of plant leaves. Our ResNet-50-based model, when used with the One-Hundred Plant Species Leaves dataset from the UCI Machine Learning Repository, achieved over 96% accuracy and demonstrated efficient computational performance. It's an incremental approach that may be used not only to watch biodiversity and agricultural diagnostics but in addition to supply a scalable basis for the implementation of ecological and visual recognition systems in the long run.

In regards to the Creator

Sherin Sunny is a Senior Engineering Manager at Walmart Vizio, where he leads the core engineering team answerable for large-scale Automatic Content Recognition (ACR) in AWS Cloud. His work spans cloud migrations, AI ML driven intelligent pipelines, vector search systems, and real-time data platforms that power next-generation content analytics.

References

[1] M. R. Popp, N. E. Zimmermann and P. Brun, Evaluating the usage of automated plant identification tools in biodiversity monitoring—a case study in Switzerland (2025), , 90, 103316.

[2] A. G. Hart, H. Bosley, C. Hooper, J. Perry, J. Sellors‐Moore, O. Moore and A. E. Goodenough, Assessing the accuracy of free automated plant identification applications (2023), , 5(3).

[3] G. Tariku, I. Ghiglieno, G. Gilioli, F. Gentilin, S. Armiraglio and I. Serina, Automated identification and classification of plant species in heterogeneous plant areas using unmanned aerial vehicle-collected RGB images and transfer learning (2023), , 7(10), 599.

[4] F. Deng, C. H. Feng, N. Gao and L. Zhang, Normalization and choosing non-differentially expressed genes improve machine learning modelling of cross-platform transcriptomic data (2025), .