Mact0/rf-detr-package-detection

RF-DETR Package Detection Model

This model is a fine-tuned RF-DETR Medium model for package/box detection, trained on a custom dataset.

Model Description

• Model Type: RF-DETR (Real-time DETR for object detection)
• Base Model: RF-DETR Medium
• Task: Object Detection (Package/Box Segmentation)
• Training Data: Custom dataset
• Classes: 2 class(es)

Training Details

• Epochs: N/A
• Batch Size: 16
• Learning Rate: 5e-05
• Input Resolution: 576x576

Performance

Training metrics available in the model repository.

Usage

Installation

pip install rfdetr torch torchvision

Loading the Model

import torch
from rfdetr import RFDETRMedium
from PIL import Image

# Load model
model = RFDETRMedium()
checkpoint = torch.load("checkpoint_best_total.pth", map_location='cpu')
model.model.load_state_dict(checkpoint['model'])
model.model.eval()

# Run inference
image = Image.open("path/to/image.jpg")
results = model.predict(image)

API Usage (with Inference Endpoints)

Once deployed as an Inference Endpoint:

import requests
from PIL import Image
import io

API_URL = "https://api-inference.huggingface.co/models/YOUR_USERNAME/rf-detr-box-segmentation"
headers = {"Authorization": "Bearer YOUR_HF_TOKEN"}

# Send image
with open("image.jpg", "rb") as f:
    data = f.read()
response = requests.post(API_URL, headers=headers, data=data)
results = response.json()

Model Details

• Developed by: Your Name
• Model date: 1762288019.2803671
• Framework: PyTorch
• License: Apache 2.0

Citation

@software{rfdetr2024,
  title = {RF-DETR: Real-time DETR},
  author = {Roboflow},
  year = {2024},
  url = {https://github.com/roboflow/rf-detr}
}

Limitations

This model is trained on a specific package detection dataset and may not generalize to all object detection tasks.