Object detection for aerial vehicles (drones, helicopters, airplanes, etc) with bounding box prediction.
The dataset we used is available on Kaggle and Roboflow.
Kaggle: https://www.kaggle.com/datasets/cybersimar08/drone-detection
Roboflow: https://universe.roboflow.com/ahmedmohsen/drone-detection-new-peksv
Using the following metrics/benchmarks, we'll be evaluating each model's ability to distinguish between aerial vehicle types and produce bounding boxes for them within the image:
- Mean Average Precision (mAP)
- Intersection over Union (IoU)
Alongside the metrics above, we'll be profiling the models to determine performance:
- Inference Latency
- Model Size
- Training Time