Yifan Wang

@jiaswee1995 Because the LISA dataset is not included in this project. You should download it from [this page](http://cvrr.ucsd.edu/vivachallenge/index.php/signs/sign-detection/). Oppps... It seems like the author closed the website :( Maybe you...

@harveenchadha From [LISA data page](http://cvrr.ucsd.edu/vivachallenge/index.php/signs/sign-detection/), you can search and find a "[Python toolkit](http://cvrr.ucsd.edu/vivachallenge/wp-content/uploads/2015/03/annotationTools.zip)". You can modify the code and get annotations in your format.

It's [AUC](http://cvrr.ucsd.edu/vivachallenge/index.php/signs/sign-detection/) is 91.88%, and the tiny model's AUC is 58.19%.

@zhji2822 I got your problem. The 6 labels (speedLimit, warning, noTurn, stop, pedestrianCrossing, signalAhead) are the superclasses of signs. They are defined in the [LISA](http://cvrr.ucsd.edu/vivachallenge/index.php/signs/sign-detection/) project papers. You can google...

@zhji2822 You should carefully read the first 10 lines of `evaluateDetections.py` and `filterAnnotationFile.py`. `[-c category]` -- `category` should be a txt file.

@zhang11wu4 You should download the pre-trained weights from the [Darknet project website](https://pjreddie.com/darknet/yolo/), and then use the cfg files to train YOLO-LISA model.

@zhji2822 This one [darknet19_448.conv.23](https://pjreddie.com/media/files/darknet19_448.conv.23) from [YOLOv2](https://pjreddie.com/darknet/yolov2/). You can read the chapter 'Training YOLO on VOC' from that page. However, [YOLOv3](https://pjreddie.com/darknet/yolo/) has been already born, so I guess using the new...