huytranduck/efficientnet_b0_50x_dataset

Model Description

This is a fine-tuned EfficientNet-B0 model for image classification with 100 classes.

The model is based on google/efficientnet-b0 and has been fine-tuned on a custom dataset for specialized image classification tasks.

Model Architecture

• Base Model: EfficientNet-B0
• Framework: PyTorch + Transformers
• Task: Image Classification
• Classes: 100
• Input Size: 224x224 pixels

Classes

['Acacia melanoxylon', 'Acer saccharinum', 'Afzelia africana', 'Afzelia pachyloba', 'Afzelia quanzensis', 'Albizia lucida (Albizia lucidior)', 'Allophylus cobbe (Pometia pinnata)', 'Anisoptera costata (Anisoptera Robusta)', 'Apuleia\xa0leiocarpa', 'Artocarpus calophyllus (Artocarpus asperulus)', 'Artocarpus heterophyllus', 'Autranella congolensis', 'Berlinia bracteosa', 'Betula pendula', 'Bobgunnia fistuloides (Swartzia fistuloides)', 'Brachystegia sp', 'Burckella obovata', 'Burretiodendron tonkinense', 'Callitris columellaris', 'Calocedrus sp', 'Canarium album', 'Chrysophyllum sp', 'Cinnamomum camphora', 'Clarisia racemosa', 'Colophospermum mopane', 'Cunninghamia lanceolata', 'Cupressus funebris (Cupressus pendula)', 'Cylicodiscus gabunensis', 'Dalbergia cochinchinensis', 'Dalbergia oliveri', 'Detarium macrocarpum', 'Dialium bipindense', 'Didelotia africana', 'Diospyros mun', 'Diospyros salletii', 'Distemonanthus benthamianus', 'Engelhardia chrysolepis (Engelhardia roxburghiana)', 'Entandrophragma cylindricum', 'Entandrophragma utile', 'Erythrophleum fordii\xa0', 'Erythrophleum ivorense', 'Eucalyptus cladocalyx', 'Eucalyptus grandis', 'Eucalyptus microcorys', 'Eucalyptus saligna', 'Fokienia hodginsii', 'Fraxinus excelsior', 'Gilbertiodendron dewevrei', 'Guarea cedrata', 'Guibourtia coleosperma', 'Heritiera littoralis', 'Hevea brasiliensis', 'Homalium caryophyllaceum', 'Homalium foetidum', 'Hopea iriana', 'Hopea pierrei', 'Hymenaea courbaril', 'Hymenolobium heterocarpum', 'Juglans regia', 'Khaya senegalensis', 'Klainedoxa gabonensis', 'Lithocarpus ducampii', 'Lophira alata', 'Magnolia hypolampra', 'Martiodendron parviflorum', 'Milicia excelsa', 'Milicia regia', 'Millettia laurentii', 'Monopetalanthus letestui (Bikinia letestui)', 'Myracrodruon urundeuva', 'Myroxylon balsamum', 'Myroxylon balsamum_v2', 'Myroxylon peruiferum', 'Nauclea diderrichii', 'Pachyelasma tessmannii', 'Palaquium waburgianum', 'Pericopsis elata', 'Pinus sp', 'Piptadeniastrum africanum', 'Populus sp', 'Prunus serotina', 'Pterocarpus macrocarpus', 'Pterocarpus soyauxii', 'Pterocarpus sp', 'Qualea paraensis', 'Quercus petraea', 'Quercus robur', 'Quercus rubra', 'Samanea saman', 'Shorea hypochra (Anthoshorea hypochra)', 'Shorea roxburghii (Anthoshorea roxburghii)', 'Sindora cochinchinensis', 'Staudtia stipitata', 'Syzygium hemisphericum (Syzygium chanlos)', 'Tarrietia cochinchinensis (Heritiera cochinchinesis)', 'Tectona grandis', 'Terminalia superba', 'Tetraberlinia bifoliolata', 'Toona sureni', 'Xylia xylocarpa']

Training Details

Training Configuration

• Epochs: 2
• Batch Size: 64
• Learning Rate: 0.0001
• Optimizer: AdamW
• Loss Function: CrossEntropyLoss
• Image Size: 224x224
• Data Augmentation: ✅ Enabled
  • Random Horizontal Flip (p=0.5)
  • Random Vertical Flip (p=0.2)
  • Random Affine (shear=0.2)
  • Color Jitter (brightness, contrast, saturation, hue=0.2)
  • Random Rotation (15°)

Data Augmentation Pipeline

transforms.Compose([
    transforms.RandomApply([
        transforms.ColorJitter(brightness=0.2, contrast=0.2, saturation=0.2, hue=0.2),
        transforms.RandomHorizontalFlip(p=0.5),
        transforms.RandomVerticalFlip(p=0.2),
        transforms.RandomAffine(degrees=15, shear=0.2, scale=(0.9, 1.1)),
        transforms.RandomRotation(degrees=15),
    ], p=0.5),
    transforms.Resize((224, 224)),
    transforms.RandomResizedCrop(224, scale=(0.8, 1.0)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])

Performance Metrics

• Metrics will be updated after evaluation

Usage

Quick Start

from transformers import EfficientNetForImageClassification, EfficientNetImageProcessor
import torch
from PIL import Image
import requests

# Load model and processor
model = EfficientNetForImageClassification.from_pretrained("huytranduck/efficientnet_b0_50x_dataset")
processor = EfficientNetImageProcessor.from_pretrained("huytranduck/efficientnet_b0_50x_dataset")

# Load image
url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)

# Preprocess image
inputs = processor(image, return_tensors="pt")

# Make prediction
with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_class_id = predictions.argmax().item()
    confidence = predictions.max().item()

print(f"Predicted class: {predicted_class_id}")
print(f"Confidence: {confidence:.4f}")

Batch Inference

# For multiple images
images = [image1, image2, image3]  # List of PIL Images
inputs = processor(images, return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_classes = predictions.argmax(dim=-1)

GPU Inference

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)

inputs = processor(image, return_tensors="pt")
inputs = {k: v.to(device) for k, v in inputs.items()}

with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)

Model Performance

The model was trained with extensive data augmentation and achieved strong performance on the validation set. The training process used:

• Weighted F1-Score for handling class imbalance
• Best model saving based on F1-Score
• AdamW optimizer with learning rate 0.0001

Training Code

Click to expand training code

import torch
from transformers import EfficientNetConfig, EfficientNetForImageClassification, EfficientNetImageProcessor

# Model setup
model_name = "google/efficientnet-b0"
config = EfficientNetConfig.from_pretrained(model_name)
config.num_labels = 100
model = EfficientNetForImageClassification(config)
processor = EfficientNetImageProcessor.from_pretrained(model_name)

# Training parameters
epochs = 2
batch_size = 64
learning_rate = 0.0001

# Training loop with F1-score monitoring
optimizer = torch.optim.AdamW(model.parameters(), lr=learning_rate)
criterion = torch.nn.CrossEntropyLoss()

Citation

If you use this model, please cite:

@misc{huytranduck_efficientnet_b0_50x_dataset,
  title={Custom EfficientNet-B0 for Image Classification},
  author={Your Name},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/huytranduck/efficientnet_b0_50x_dataset}
}

License

This model is released under the Apache 2.0 License.