Bounding box is not right.

[HSHunterR]

I try on my dataset(360 images of 3 class objects),modify model tiny-yolo-4c, batch 8, use the default lr, gpu model(GTX 950M 2GB), about 9k steps(more than 10 hours), loss about 3 or 2, move ave loss the same, but the bounding box is not correct on train image( threshold is 0.0045); And then I use 6 images, train 150 epoch, batch 6, the loss is 3 or 2, then it can't more smaller, the bounding box is not correct too, how to solve it?

[kmarrip]

@HSHunterR loss is too high for any bounding box to be true , add more images to your dataset, for each class have at least 400 images, then retrain the model, make sure the loss is around 0.5

[HSHunterR]

Thanks for your suggestion！