One of the challenge we face is to improve the accuracy of the object detection model. We are evaluating the trained model with different setting on the architecture and hyperparameter. We have also examined the improvement by the image pre-processing method like histogram equalization.
LI, Chun Ming
January 19, 2022 3:03 pm
Nice poster design with detailed descriptions! The methodology and system flow part are both intuitive and very straightforward. For the web application development part, I suggest “Express.js” with “React” as it is commonly used to implement the front-end(UI) and backend for using APIs.
Yes, it is based on visible light. We have collect our own datasets and photos are collected both in day time and night time, it will have difference since light condition is an important issue in object detection. Some of the photos we got are in very dark situation and it is hard to detect the car with the bounding box. But it is hard to overcome it since it is the limitation of the camera. Thank you for your question and I hope I can answer it.
First of all, thank you for your question! The answer of the first question is yes. The object detection system are using the monocular image to detect the vehicle and hence it is based on visible light; We observe that the qualitative result of the system drops comparing to daytime when we test our system using self-collected testing datasets that includes the night and daytime.
what’s the main challenge in your project?
One of the challenge we face is to improve the accuracy of the object detection model. We are evaluating the trained model with different setting on the architecture and hyperparameter. We have also examined the improvement by the image pre-processing method like histogram equalization.
Nice poster design with detailed descriptions! The methodology and system flow part are both intuitive and very straightforward. For the web application development part, I suggest “Express.js” with “React” as it is commonly used to implement the front-end(UI) and backend for using APIs.
Thank you for your comment and suggestion. We will take them for our consideration.
Is the object detection system base on visible light? Will the efficiency or accuracy of the system differ in night compare to daytime?
Yes, it is based on visible light. We have collect our own datasets and photos are collected both in day time and night time, it will have difference since light condition is an important issue in object detection. Some of the photos we got are in very dark situation and it is hard to detect the car with the bounding box. But it is hard to overcome it since it is the limitation of the camera. Thank you for your question and I hope I can answer it.
First of all, thank you for your question! The answer of the first question is yes. The object detection system are using the monocular image to detect the vehicle and hence it is based on visible light; We observe that the qualitative result of the system drops comparing to daytime when we test our system using self-collected testing datasets that includes the night and daytime.
Nice poster!
Thank you so much