Eric Liu

The bdd2coco.py will split json files one by one , the message you showed "numbers of images: 2" , it is very strange. It should above 60k images

Sorry for late reply , the step2 output will split json file to each image annotation, which was just like MS-COCO dataset , and I just upload the step 3...

Sorry all , I forgot the split annotation steps , and I just upload it . @dishita26 the train and val list file will be generated at spilt annotation step

@dishita26 I uploaded val lmdb [here](https://drive.google.com/open?id=1MI0KzphTY5a1wijXvJua3X4X0woLrSUw)

@sainisanjay Just modify the label to background in [prototxt](https://github.com/eric612/MobileNet-YOLO/blob/master/data/bdd100k/labelmap_detection.prototxt), and then reproduce lmdb . For ex, set "traffic sign" label to zero .

The original implement was come from here : https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md The inference time of v2 was 27 ms and v1 was 31 ms , but it is base on tensorflow framework...

I just update the demo_webcam.cmd at latest version. It would help to you.

Yes , segmentation task only has two classes which were [main lane and other lanes](https://www.youtube.com/watch?v=O6kkRlT0UWE&ab_channel=Zegapain)

For example , in my training script , just replace mobilenet_iter_73000.caffemodel to your specify caffmodel ``` ./tools/caffe train --solver models/yolov3/mobilenet_yolov3_lite_solver.prototxt --weights models/MobileNet/mobilenet_iter_73000.caffemodel --gpu=0 2>&1 | tee $LOG ```

Thanks for your feedback , I just update a new pva deploy prototxt