This repository helps in development and training of YOLOv3 network for real-time detection of small objects such as traffic signs.
This repository is licensed under MIT license. This work is adaption from AntonMu/TrainYourOwnYOLO, which was inspired from qqwweee/keras-yolo3
[Hafsa Amanullah] - 🌐 Github - 🌐 LinkedIn Profile
[Prof. Dr. Min Young Kim] - 🌐 Google scholar
[Dr. Yawar Rehman] - 🌐 Github - 🌐 LinkedIn Profile
You may use this code for small traffic sign detection by following these simple steps.
1- Estimate anchors using anchors.mat file
2- Copy test and train images in Data/Source_images/Test_images and Data/Source_images/Training_images respectively
3- Use Train_YOLO.py to train your network.
4- Use Detector.py to test the trained network with test images.
The remaining codes will be uploaded soon
Traffic sign detection in GTSDB dataset (a) traffic sign detection with size variation (b) small traffic sign detection (c) a larger traffic sign recognition (d) small traffic sign detection
Traffic sign detection in STS dataset (a-c) small-sized traffic sign detection (d) a large-sized traffic sign detection
Please cite the following if this code/work is helpful to you
- Y. Rehman, H. Amanullah, M. A. Shirazi and M. Y. Kim, "Small Traffic Sign Detection in Big Images: Searching Needle in a Hay," in IEEE Access, vol. 10, pp. 18667-18680, 2022. link
- Rehman, Y.; Amanullah, H.; Saqib Bhatti, D.M.; Toor, W.T.; Ahmad, M.; Mazzara, M. Detection of Small Size Traffic Signs Using Regressive Anchor Box Selection and DBL Layer Tweaking in YOLOv3. Appl. Sci. 2021, 11, 11555. link