This project focuses on detecting damaged packages using a custom-trained YOLOv5 model. The goal is to identify packages with visible signs of damage in real-time, aiding in the automation of package inspection in logistics operations.
- dataset: https://pan.baidu.com/s/1JKo1WqwrIbdH88Zz6ZJmjg?pwd=35v0 code：35v0
- best.pt: Includes the YOLOv5 model
-
Data Collection: We gathered a dataset of package images, ensuring a balanced representation of both damaged and undamaged packages.
-
Data Preprocessing: The images were resized, and labels were annotated using YOLO format. Data augmentation techniques were applied to increase the diversity of the training set.
-
Model Training:
- The YOLOv5 model was trained on the preprocessed dataset.
- Hyperparameters were tuned to optimize model performance, including learning rate, batch size, and number of epochs.
-
Evaluation: The model was evaluated on a separate validation set. Metrics such as precision, recall, and mAP (mean Average Precision) were calculated to assess the model's performance.
-
Inference: The trained model was used to detect damaged packages in new images. The results were visualized and saved for further analysis.
- Clone the Repository:
git clone https://github.com/yourusername/damaged-package-detection.git cd damaged-package-detection