Vincent Liu
Hi @meijieru , Since lmdb stores the data in the dictionary, the sequences based on length is not preserved. Any ideas?
Check the annotation. There may be some annotated boxes without class name "Person008" which does not match the parameter _object_names_array_ in _setTrainConfig()_
Any live link to V2?
May I know if the current link is https://imglab.in/ ? I tried but still got floating point in the dlib xml. Any ideas?
Hi @princewang1994 , I found that these files have 3 vertices only polygons which cause the scripts failed: poly_gt_img724.mat poly_gt_img718.mat poly_gt_img1301.mat poly_gt_img1075.mat poly_gt_img914.mat poly_gt_img914.mat poly_gt_img1474.mat poly_gt_img912.mat poly_gt_img1304.mat poly_gt_img1489.mat poly_gt_img720.mat poly_gt_img1160.mat...
Hi, @15926273249 我沒有細究這個issue的原因，不過換了pytorch 的版本就沒有這個問題了。以下是我的安裝指令，給您參考： pip install torch==1.5.0+cu101 torchvision==0.6.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html
> @mrljwlm 👋 Hello! Thanks for asking about **resuming training**. YOLO 🚀 Learning Rate (LR) [schedulers](https://github.com/ultralytics/yolov5/blob/aa542ce6a65658ff931fee9bbab77c0145c152f0/train.py#L175-L181) follow predefined LR curves for the fixed number of `epochs` defined at training start...
Same here, anyone has the solution please
Yeah, i think we need to modify the network for other shape. However, according to the readme, we can train directly with height=64?