Open MTzhong opened 1 year ago
Part of train images
Our YOLOv3 was trained on the COWC-M dataset, which comprises of aerial images of cars. To train a patch, we collected and used aerial images of cars similar to those in the COWC-M dataset i.e., cars in COWC-M, Side Street, and Car Park are of similar size.
Based on your images above, it seems like you want to train a patch on earth based images of cars. If so, you will need to train YOLOv3 on those types of images.
Hello , I try to train my own patch using my own dataset, but I get a patch that doesn't work. I can get a good patch using your dataset. So I am confused with this problem and where should i make changes in this code. Thanks.