Project Details
- Developed an object detection system using the YOLOv3 deep learning algorithm
- Utilized Python and OpenCV libraries to implement the algorithm
- Researched and analyzed various deep learning algorithms before settling on YOLOv3
- Fine-tuned YOLO algorithm using Tensorflow v2.0
- Utilized transfer learning techniques to improve the accuracy of the model
- Trained the model on a large dataset of images and annotations
- Incorporated pre-trained weights from the COCO dataset to improve the model's accuracy.
- COCO dataset contains 80 classes of objects and is widely used for object detection tasks
- Performed hyperparameter tuning to optimize the model's performance.
- Experimented with different learning rates, batch sizes, and activation functions
- Monitored the model's performance on a validation set to avoid overfitting
- Implemented non-maximum suppression to improve the model's ability to detect multiple
instances of an object.- Non-maximum suppression is a technique used to eliminate redundant detections
- Improved the overall accuracy of the model by reducing false positive detections
- Evaluated the model using metrics such as mean average precision (MAP) and Intersection
over Union (IoU)- MAP is a measure of the model's ability to correctly detect objects in an image
- IoU measures the overlap between the ground truth and predicted bounding boxes
- Focused on the cloud platform Google Colab for the implementation
- Utilized Google Colab's GPU to speed up the training process
- Avoided the need for expensive hardware by leveraging cloud computing resources
- Finished in first place during the department-wide symposium
- Presented the project to a panel of experts and received high praise for its accuracy and performance.
- Demonstrated the ability to apply deep learning concepts in a practical setting.
Images: