frabob2017

> @frabob2017 user error. Colab instances never provide more than a single GPU, yet you've requested 2. Just follow the default examples in the official notebook: https://colab.research.google.com/github/ultralytics/yolov5/blob/master/tutorial.ipynb Good to know...

> @frabob2017 👋 Hello! Thanks for asking about **training speed issues**. YOLOv5 🚀 can be trained on CPU (slowest), [single-GPU](https://docs.ultralytics.com/yolov5/tutorials/train_custom_data), or [multi-GPU](https://docs.ultralytics.com/yolov5/tutorials/multi_gpu_training) (fastest). If you would like to increase your...

> @frabob2017 This is very basic: images are normally compressed when saved on disk. I though there is a configure file that can control the data format such as reducing...

Hello Martin, Yes, it is the val_batch1_labels.jpg, val_batch2_labels.jpg; val_batch3_labels.jpg. It is the labels program read from the ground truth labels. The position of bounding box shift. Frank

Hello Martin, Thank you for your testing. You see, right side one is correct. It covers the fracture area. The left side one is NOT correct. I used Yolo format....

Radiologist coordinate system is not normalized. What I did is normalized them. The image size is 512, 512; what I did is just normalized by 512.

Hello Martin, Thank you so much for help. I double check again PIL.rectangle(x,y), x,y is left bottom corner. I make a mistake and think it is center point. Thank you...

this problem is solved. thank you so much.

> @frabob2017 not sure I understand your question. To update hyperparameters just update hyperparameters: > > https://github.com/ultralytics/yolov5/blob/11640698977724daf7982c9da398c2ee2f2b6e91/data/hyps/hyp.scratch-low.yaml#L25 > @frabob2017 not sure I understand your question. To update hyperparameters just update...

> @frabob2017 I would just reduce your background images and train normally. it is impossible for my task to reduce too many background images. I want to confirm that too...