InfraOD Model

Infrared Object Detection with YOLOv11 Nano

---

🚀 Overview

This folder contains the trained YOLOv11 nano model for infrared object detection. The model was trained and tested on over 35,000 images from diverse scenarios and camera angles, enabling robust detection of infrared objects.

---

📊 Model & Training Details (click to expand)

• Model: YOLOv11 nano (yolo11n.pt)
• Task: Object Detection
• Images: 35,000+ (train/test split)
• Epochs: 50
• Batch Size: 16
• Image Size: 336x336
• Augmentations: RandAugment, Mosaic, Flip, Erasing, etc.
• Framework: PyTorch
• Config: See args.yaml

---

🏆 Results & Visualizations

Below are some of the key results and visualizations generated during model evaluation:

Confusion Matrix	Normalized Confusion Matrix	Results Curve

Sample Predictions

Validation Batch 0	Validation Batch 1	Validation Batch 2

---

📈 Training Metrics (click to expand)

• See results.csv for detailed epoch-wise metrics (loss, precision, recall, mAP, etc.).
• Example (final epoch):
  • Precision: 0.796
  • Recall: 0.614
  • mAP@0.5: 0.690
  • mAP@0.5:0.95: 0.493

---

🧩 Model Weights

• Best Model: weights/best.pt
• Last Epoch Model: weights/last.pt
• ONNX Export: weights/best.onnx

Note: These files are large. Download as needed for inference or further training.

---

🛠️ How to Use

1. Clone the repository:

   git clone https://github.com/pranayjoshi/InfraOD.git
   cd InfraOD/InfraOD_model

2. Download model weights (if not present):
   • Place them in the weights/ directory.
3. Run inference:
   • Use your preferred YOLOv11 inference script, specifying the model path (e.g., weights/best.pt).

---

📚 Credits & References

• YOLOv11 nano
• HIT-UAV Infrared Thermal Dataset
• LLVIP
• Kaggle: Thermal Images for Human Detection
• Original Project Inspiration

---

For questions or contributions, please open an issue or pull request!