You can deploy this model and start earning money today!

johngunerli/
fasterrcnn-package-detector

Name: fasterrcnn-package-detector
Availability: InStock

Views

No views yet

Faster R-CNN Package Detector

This repository contains a Faster R-CNN model trained to detect packages at front doors using the Roboflow dataset. The project includes dataset downloading, model training, evaluation, and deployment to the Hugging Face Model Hub.

📌 Features

📂 Dataset Download: Uses Roboflow API to fetch the dataset in YOLO format.
🏋️ Model Training: Implements Faster R-CNN with ResNet-50 backbone for object detection.
🏆 Evaluation: Computes Intersection over Union (IoU) to measure model accuracy.
☁️ Model Deployment: Uploads trained model to Hugging Face Hub for easy sharing and reuse.

🚀 Installation

1️⃣ Setup Environment

Ensure you have Python 3 installed, and set up dependencies using:

pip install torch torchvision roboflow tqdm huggingface_hub

2️⃣ Clone the Repository

bash

git clone https://huggingface.co/johngunerli/fasterrcnn-package-detector
cd fasterrcnn-package-detector

📊 Dataset

The dataset is fetched from Roboflow using:

python

from roboflow import Roboflow

rf = Roboflow(api_key="YOUR_API_KEY")
project = rf.workspace("package-detection").project("package-at-front-door")
version = project.version(2)
dataset = version.download("yolov5")

🏋️ Training

To train the model on the downloaded dataset, run:

python train_model.py

Training Configuration

Backbone: ResNet-50
Optimizer: SGD (lr=0.005, momentum=0.9, weight_decay=0.0005)
Scheduler: StepLR (step_size=3, gamma=0.1)
Epochs: 10

🎯 Evaluation

The model evaluates performance using IoU scores:

python

val_loss, avg_iou = evaluate_model(model, valid_loader, device)
print(f"Validation Loss: {val_loss:.4f}, Avg IoU: {avg_iou:.4f}")

☁️ Uploading to Hugging Face

To upload the trained model to Hugging Face, run:

python

from huggingface_hub import Repository

repo_id = "your-username/fasterrcnn-package-detector"
repo_dir = "./fasterrcnn-package-detector"

repo = Repository(repo_dir, clone_from=repo_id)

torch.save(model.state_dict(), "model_package.pth")
repo.push_to_hub(commit_message="Upload Faster R-CNN trained model")

📌 Model Performance

Epoch	Train Loss	Avg IoU
1	146.32	0.7974
2	123.80	0.7757
3	117.00	0.7855
4	107.56	0.8124
5	107.14	0.8129
6	103.35	0.8012
7	102.27	0.8132
8	102.88	0.8197
9	100.50	0.8244
10	103.23	0.8214

📌 Usage

To use the trained model for inference:

python

import torch
import torchvision
from torchvision.models.detection.faster_rcnn import FastRCNNPredictor

def get_model(num_classes):
    model = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=False)
    in_features = model.roi_heads.box_predictor.cls_score.in_features
    model.roi_heads.box_predictor = FastRCNNPredictor(in_features, num_classes)
    return model

model = get_model(2)
model.load_state_dict(torch.load("model_package.pth", map_location="cpu"))
model.eval()